1/*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_shared.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_mount.h"
25#include "xfs_inode.h"
26#include "xfs_ioctl.h"
27#include "xfs_alloc.h"
28#include "xfs_rtalloc.h"
29#include "xfs_itable.h"
30#include "xfs_error.h"
31#include "xfs_attr.h"
32#include "xfs_bmap.h"
33#include "xfs_bmap_util.h"
34#include "xfs_fsops.h"
35#include "xfs_discard.h"
36#include "xfs_quota.h"
37#include "xfs_export.h"
38#include "xfs_trace.h"
39#include "xfs_icache.h"
40#include "xfs_symlink.h"
41#include "xfs_trans.h"
42#include "xfs_pnfs.h"
43#include "xfs_acl.h"
44
45#include <linux/capability.h>
46#include <linux/dcache.h>
47#include <linux/mount.h>
48#include <linux/namei.h>
49#include <linux/pagemap.h>
50#include <linux/slab.h>
51#include <linux/exportfs.h>
52
53/*
54 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
55 * a file or fs handle.
56 *
57 * XFS_IOC_PATH_TO_FSHANDLE
58 *    returns fs handle for a mount point or path within that mount point
59 * XFS_IOC_FD_TO_HANDLE
60 *    returns full handle for a FD opened in user space
61 * XFS_IOC_PATH_TO_HANDLE
62 *    returns full handle for a path
63 */
64int
65xfs_find_handle(
66	unsigned int		cmd,
67	xfs_fsop_handlereq_t	*hreq)
68{
69	int			hsize;
70	xfs_handle_t		handle;
71	struct inode		*inode;
72	struct fd		f = {NULL};
73	struct path		path;
74	int			error;
75	struct xfs_inode	*ip;
76
77	if (cmd == XFS_IOC_FD_TO_HANDLE) {
78		f = fdget(hreq->fd);
79		if (!f.file)
80			return -EBADF;
81		inode = file_inode(f.file);
82	} else {
83		error = user_lpath((const char __user *)hreq->path, &path);
84		if (error)
85			return error;
86		inode = d_inode(path.dentry);
87	}
88	ip = XFS_I(inode);
89
90	/*
91	 * We can only generate handles for inodes residing on a XFS filesystem,
92	 * and only for regular files, directories or symbolic links.
93	 */
94	error = -EINVAL;
95	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
96		goto out_put;
97
98	error = -EBADF;
99	if (!S_ISREG(inode->i_mode) &&
100	    !S_ISDIR(inode->i_mode) &&
101	    !S_ISLNK(inode->i_mode))
102		goto out_put;
103
104
105	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
106
107	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
108		/*
109		 * This handle only contains an fsid, zero the rest.
110		 */
111		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
112		hsize = sizeof(xfs_fsid_t);
113	} else {
114		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
115					sizeof(handle.ha_fid.fid_len);
116		handle.ha_fid.fid_pad = 0;
117		handle.ha_fid.fid_gen = ip->i_d.di_gen;
118		handle.ha_fid.fid_ino = ip->i_ino;
119
120		hsize = XFS_HSIZE(handle);
121	}
122
123	error = -EFAULT;
124	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
125	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
126		goto out_put;
127
128	error = 0;
129
130 out_put:
131	if (cmd == XFS_IOC_FD_TO_HANDLE)
132		fdput(f);
133	else
134		path_put(&path);
135	return error;
136}
137
138/*
139 * No need to do permission checks on the various pathname components
140 * as the handle operations are privileged.
141 */
142STATIC int
143xfs_handle_acceptable(
144	void			*context,
145	struct dentry		*dentry)
146{
147	return 1;
148}
149
150/*
151 * Convert userspace handle data into a dentry.
152 */
153struct dentry *
154xfs_handle_to_dentry(
155	struct file		*parfilp,
156	void __user		*uhandle,
157	u32			hlen)
158{
159	xfs_handle_t		handle;
160	struct xfs_fid64	fid;
161
162	/*
163	 * Only allow handle opens under a directory.
164	 */
165	if (!S_ISDIR(file_inode(parfilp)->i_mode))
166		return ERR_PTR(-ENOTDIR);
167
168	if (hlen != sizeof(xfs_handle_t))
169		return ERR_PTR(-EINVAL);
170	if (copy_from_user(&handle, uhandle, hlen))
171		return ERR_PTR(-EFAULT);
172	if (handle.ha_fid.fid_len !=
173	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
174		return ERR_PTR(-EINVAL);
175
176	memset(&fid, 0, sizeof(struct fid));
177	fid.ino = handle.ha_fid.fid_ino;
178	fid.gen = handle.ha_fid.fid_gen;
179
180	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
181			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
182			xfs_handle_acceptable, NULL);
183}
184
185STATIC struct dentry *
186xfs_handlereq_to_dentry(
187	struct file		*parfilp,
188	xfs_fsop_handlereq_t	*hreq)
189{
190	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
191}
192
193int
194xfs_open_by_handle(
195	struct file		*parfilp,
196	xfs_fsop_handlereq_t	*hreq)
197{
198	const struct cred	*cred = current_cred();
199	int			error;
200	int			fd;
201	int			permflag;
202	struct file		*filp;
203	struct inode		*inode;
204	struct dentry		*dentry;
205	fmode_t			fmode;
206	struct path		path;
207
208	if (!capable(CAP_SYS_ADMIN))
209		return -EPERM;
210
211	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
212	if (IS_ERR(dentry))
213		return PTR_ERR(dentry);
214	inode = d_inode(dentry);
215
216	/* Restrict xfs_open_by_handle to directories & regular files. */
217	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
218		error = -EPERM;
219		goto out_dput;
220	}
221
222#if BITS_PER_LONG != 32
223	hreq->oflags |= O_LARGEFILE;
224#endif
225
226	permflag = hreq->oflags;
227	fmode = OPEN_FMODE(permflag);
228	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
229	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
230		error = -EPERM;
231		goto out_dput;
232	}
233
234	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
235		error = -EACCES;
236		goto out_dput;
237	}
238
239	/* Can't write directories. */
240	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
241		error = -EISDIR;
242		goto out_dput;
243	}
244
245	fd = get_unused_fd_flags(0);
246	if (fd < 0) {
247		error = fd;
248		goto out_dput;
249	}
250
251	path.mnt = parfilp->f_path.mnt;
252	path.dentry = dentry;
253	filp = dentry_open(&path, hreq->oflags, cred);
254	dput(dentry);
255	if (IS_ERR(filp)) {
256		put_unused_fd(fd);
257		return PTR_ERR(filp);
258	}
259
260	if (S_ISREG(inode->i_mode)) {
261		filp->f_flags |= O_NOATIME;
262		filp->f_mode |= FMODE_NOCMTIME;
263	}
264
265	fd_install(fd, filp);
266	return fd;
267
268 out_dput:
269	dput(dentry);
270	return error;
271}
272
273int
274xfs_readlink_by_handle(
275	struct file		*parfilp,
276	xfs_fsop_handlereq_t	*hreq)
277{
278	struct dentry		*dentry;
279	__u32			olen;
280	void			*link;
281	int			error;
282
283	if (!capable(CAP_SYS_ADMIN))
284		return -EPERM;
285
286	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
287	if (IS_ERR(dentry))
288		return PTR_ERR(dentry);
289
290	/* Restrict this handle operation to symlinks only. */
291	if (!d_is_symlink(dentry)) {
292		error = -EINVAL;
293		goto out_dput;
294	}
295
296	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
297		error = -EFAULT;
298		goto out_dput;
299	}
300
301	link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
302	if (!link) {
303		error = -ENOMEM;
304		goto out_dput;
305	}
306
307	error = xfs_readlink(XFS_I(d_inode(dentry)), link);
308	if (error)
309		goto out_kfree;
310	error = readlink_copy(hreq->ohandle, olen, link);
311	if (error)
312		goto out_kfree;
313
314 out_kfree:
315	kfree(link);
316 out_dput:
317	dput(dentry);
318	return error;
319}
320
321int
322xfs_set_dmattrs(
323	xfs_inode_t     *ip,
324	u_int		evmask,
325	u_int16_t	state)
326{
327	xfs_mount_t	*mp = ip->i_mount;
328	xfs_trans_t	*tp;
329	int		error;
330
331	if (!capable(CAP_SYS_ADMIN))
332		return -EPERM;
333
334	if (XFS_FORCED_SHUTDOWN(mp))
335		return -EIO;
336
337	tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
338	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
339	if (error) {
340		xfs_trans_cancel(tp);
341		return error;
342	}
343	xfs_ilock(ip, XFS_ILOCK_EXCL);
344	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
345
346	ip->i_d.di_dmevmask = evmask;
347	ip->i_d.di_dmstate  = state;
348
349	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
350	error = xfs_trans_commit(tp);
351
352	return error;
353}
354
355STATIC int
356xfs_fssetdm_by_handle(
357	struct file		*parfilp,
358	void			__user *arg)
359{
360	int			error;
361	struct fsdmidata	fsd;
362	xfs_fsop_setdm_handlereq_t dmhreq;
363	struct dentry		*dentry;
364
365	if (!capable(CAP_MKNOD))
366		return -EPERM;
367	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
368		return -EFAULT;
369
370	error = mnt_want_write_file(parfilp);
371	if (error)
372		return error;
373
374	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
375	if (IS_ERR(dentry)) {
376		mnt_drop_write_file(parfilp);
377		return PTR_ERR(dentry);
378	}
379
380	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
381		error = -EPERM;
382		goto out;
383	}
384
385	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
386		error = -EFAULT;
387		goto out;
388	}
389
390	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
391				 fsd.fsd_dmstate);
392
393 out:
394	mnt_drop_write_file(parfilp);
395	dput(dentry);
396	return error;
397}
398
399STATIC int
400xfs_attrlist_by_handle(
401	struct file		*parfilp,
402	void			__user *arg)
403{
404	int			error = -ENOMEM;
405	attrlist_cursor_kern_t	*cursor;
406	xfs_fsop_attrlist_handlereq_t al_hreq;
407	struct dentry		*dentry;
408	char			*kbuf;
409
410	if (!capable(CAP_SYS_ADMIN))
411		return -EPERM;
412	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
413		return -EFAULT;
414	if (al_hreq.buflen < sizeof(struct attrlist) ||
415	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
416		return -EINVAL;
417
418	/*
419	 * Reject flags, only allow namespaces.
420	 */
421	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
422		return -EINVAL;
423
424	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
425	if (IS_ERR(dentry))
426		return PTR_ERR(dentry);
427
428	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
429	if (!kbuf)
430		goto out_dput;
431
432	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
433	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
434					al_hreq.flags, cursor);
435	if (error)
436		goto out_kfree;
437
438	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
439		error = -EFAULT;
440
441out_kfree:
442	kmem_free(kbuf);
443out_dput:
444	dput(dentry);
445	return error;
446}
447
448int
449xfs_attrmulti_attr_get(
450	struct inode		*inode,
451	unsigned char		*name,
452	unsigned char		__user *ubuf,
453	__uint32_t		*len,
454	__uint32_t		flags)
455{
456	unsigned char		*kbuf;
457	int			error = -EFAULT;
458
459	if (*len > XFS_XATTR_SIZE_MAX)
460		return -EINVAL;
461	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
462	if (!kbuf)
463		return -ENOMEM;
464
465	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
466	if (error)
467		goto out_kfree;
468
469	if (copy_to_user(ubuf, kbuf, *len))
470		error = -EFAULT;
471
472out_kfree:
473	kmem_free(kbuf);
474	return error;
475}
476
477int
478xfs_attrmulti_attr_set(
479	struct inode		*inode,
480	unsigned char		*name,
481	const unsigned char	__user *ubuf,
482	__uint32_t		len,
483	__uint32_t		flags)
484{
485	unsigned char		*kbuf;
486	int			error;
487
488	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
489		return -EPERM;
490	if (len > XFS_XATTR_SIZE_MAX)
491		return -EINVAL;
492
493	kbuf = memdup_user(ubuf, len);
494	if (IS_ERR(kbuf))
495		return PTR_ERR(kbuf);
496
497	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
498	if (!error)
499		xfs_forget_acl(inode, name, flags);
500	kfree(kbuf);
501	return error;
502}
503
504int
505xfs_attrmulti_attr_remove(
506	struct inode		*inode,
507	unsigned char		*name,
508	__uint32_t		flags)
509{
510	int			error;
511
512	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
513		return -EPERM;
514	error = xfs_attr_remove(XFS_I(inode), name, flags);
515	if (!error)
516		xfs_forget_acl(inode, name, flags);
517	return error;
518}
519
520STATIC int
521xfs_attrmulti_by_handle(
522	struct file		*parfilp,
523	void			__user *arg)
524{
525	int			error;
526	xfs_attr_multiop_t	*ops;
527	xfs_fsop_attrmulti_handlereq_t am_hreq;
528	struct dentry		*dentry;
529	unsigned int		i, size;
530	unsigned char		*attr_name;
531
532	if (!capable(CAP_SYS_ADMIN))
533		return -EPERM;
534	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
535		return -EFAULT;
536
537	/* overflow check */
538	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
539		return -E2BIG;
540
541	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
542	if (IS_ERR(dentry))
543		return PTR_ERR(dentry);
544
545	error = -E2BIG;
546	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
547	if (!size || size > 16 * PAGE_SIZE)
548		goto out_dput;
549
550	ops = memdup_user(am_hreq.ops, size);
551	if (IS_ERR(ops)) {
552		error = PTR_ERR(ops);
553		goto out_dput;
554	}
555
556	error = -ENOMEM;
557	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
558	if (!attr_name)
559		goto out_kfree_ops;
560
561	error = 0;
562	for (i = 0; i < am_hreq.opcount; i++) {
563		ops[i].am_error = strncpy_from_user((char *)attr_name,
564				ops[i].am_attrname, MAXNAMELEN);
565		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
566			error = -ERANGE;
567		if (ops[i].am_error < 0)
568			break;
569
570		switch (ops[i].am_opcode) {
571		case ATTR_OP_GET:
572			ops[i].am_error = xfs_attrmulti_attr_get(
573					d_inode(dentry), attr_name,
574					ops[i].am_attrvalue, &ops[i].am_length,
575					ops[i].am_flags);
576			break;
577		case ATTR_OP_SET:
578			ops[i].am_error = mnt_want_write_file(parfilp);
579			if (ops[i].am_error)
580				break;
581			ops[i].am_error = xfs_attrmulti_attr_set(
582					d_inode(dentry), attr_name,
583					ops[i].am_attrvalue, ops[i].am_length,
584					ops[i].am_flags);
585			mnt_drop_write_file(parfilp);
586			break;
587		case ATTR_OP_REMOVE:
588			ops[i].am_error = mnt_want_write_file(parfilp);
589			if (ops[i].am_error)
590				break;
591			ops[i].am_error = xfs_attrmulti_attr_remove(
592					d_inode(dentry), attr_name,
593					ops[i].am_flags);
594			mnt_drop_write_file(parfilp);
595			break;
596		default:
597			ops[i].am_error = -EINVAL;
598		}
599	}
600
601	if (copy_to_user(am_hreq.ops, ops, size))
602		error = -EFAULT;
603
604	kfree(attr_name);
605 out_kfree_ops:
606	kfree(ops);
607 out_dput:
608	dput(dentry);
609	return error;
610}
611
612int
613xfs_ioc_space(
614	struct xfs_inode	*ip,
615	struct inode		*inode,
616	struct file		*filp,
617	int			ioflags,
618	unsigned int		cmd,
619	xfs_flock64_t		*bf)
620{
621	struct iattr		iattr;
622	enum xfs_prealloc_flags	flags = 0;
623	uint			iolock = XFS_IOLOCK_EXCL;
624	int			error;
625
626	/*
627	 * Only allow the sys admin to reserve space unless
628	 * unwritten extents are enabled.
629	 */
630	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
631	    !capable(CAP_SYS_ADMIN))
632		return -EPERM;
633
634	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
635		return -EPERM;
636
637	if (!(filp->f_mode & FMODE_WRITE))
638		return -EBADF;
639
640	if (!S_ISREG(inode->i_mode))
641		return -EINVAL;
642
643	if (filp->f_flags & O_DSYNC)
644		flags |= XFS_PREALLOC_SYNC;
645	if (ioflags & XFS_IO_INVIS)
646		flags |= XFS_PREALLOC_INVISIBLE;
647
648	error = mnt_want_write_file(filp);
649	if (error)
650		return error;
651
652	xfs_ilock(ip, iolock);
653	error = xfs_break_layouts(inode, &iolock, false);
654	if (error)
655		goto out_unlock;
656
657	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
658	iolock |= XFS_MMAPLOCK_EXCL;
659
660	switch (bf->l_whence) {
661	case 0: /*SEEK_SET*/
662		break;
663	case 1: /*SEEK_CUR*/
664		bf->l_start += filp->f_pos;
665		break;
666	case 2: /*SEEK_END*/
667		bf->l_start += XFS_ISIZE(ip);
668		break;
669	default:
670		error = -EINVAL;
671		goto out_unlock;
672	}
673
674	/*
675	 * length of <= 0 for resv/unresv/zero is invalid.  length for
676	 * alloc/free is ignored completely and we have no idea what userspace
677	 * might have set it to, so set it to zero to allow range
678	 * checks to pass.
679	 */
680	switch (cmd) {
681	case XFS_IOC_ZERO_RANGE:
682	case XFS_IOC_RESVSP:
683	case XFS_IOC_RESVSP64:
684	case XFS_IOC_UNRESVSP:
685	case XFS_IOC_UNRESVSP64:
686		if (bf->l_len <= 0) {
687			error = -EINVAL;
688			goto out_unlock;
689		}
690		break;
691	default:
692		bf->l_len = 0;
693		break;
694	}
695
696	if (bf->l_start < 0 ||
697	    bf->l_start > inode->i_sb->s_maxbytes ||
698	    bf->l_start + bf->l_len < 0 ||
699	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
700		error = -EINVAL;
701		goto out_unlock;
702	}
703
704	switch (cmd) {
705	case XFS_IOC_ZERO_RANGE:
706		flags |= XFS_PREALLOC_SET;
707		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
708		break;
709	case XFS_IOC_RESVSP:
710	case XFS_IOC_RESVSP64:
711		flags |= XFS_PREALLOC_SET;
712		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
713						XFS_BMAPI_PREALLOC);
714		break;
715	case XFS_IOC_UNRESVSP:
716	case XFS_IOC_UNRESVSP64:
717		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
718		break;
719	case XFS_IOC_ALLOCSP:
720	case XFS_IOC_ALLOCSP64:
721	case XFS_IOC_FREESP:
722	case XFS_IOC_FREESP64:
723		flags |= XFS_PREALLOC_CLEAR;
724		if (bf->l_start > XFS_ISIZE(ip)) {
725			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
726					bf->l_start - XFS_ISIZE(ip), 0);
727			if (error)
728				goto out_unlock;
729		}
730
731		iattr.ia_valid = ATTR_SIZE;
732		iattr.ia_size = bf->l_start;
733
734		error = xfs_setattr_size(ip, &iattr);
735		break;
736	default:
737		ASSERT(0);
738		error = -EINVAL;
739	}
740
741	if (error)
742		goto out_unlock;
743
744	error = xfs_update_prealloc_flags(ip, flags);
745
746out_unlock:
747	xfs_iunlock(ip, iolock);
748	mnt_drop_write_file(filp);
749	return error;
750}
751
752STATIC int
753xfs_ioc_bulkstat(
754	xfs_mount_t		*mp,
755	unsigned int		cmd,
756	void			__user *arg)
757{
758	xfs_fsop_bulkreq_t	bulkreq;
759	int			count;	/* # of records returned */
760	xfs_ino_t		inlast;	/* last inode number */
761	int			done;
762	int			error;
763
764	/* done = 1 if there are more stats to get and if bulkstat */
765	/* should be called again (unused here, but used in dmapi) */
766
767	if (!capable(CAP_SYS_ADMIN))
768		return -EPERM;
769
770	if (XFS_FORCED_SHUTDOWN(mp))
771		return -EIO;
772
773	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
774		return -EFAULT;
775
776	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
777		return -EFAULT;
778
779	if ((count = bulkreq.icount) <= 0)
780		return -EINVAL;
781
782	if (bulkreq.ubuffer == NULL)
783		return -EINVAL;
784
785	if (cmd == XFS_IOC_FSINUMBERS)
786		error = xfs_inumbers(mp, &inlast, &count,
787					bulkreq.ubuffer, xfs_inumbers_fmt);
788	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
789		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
790					sizeof(xfs_bstat_t), NULL, &done);
791	else	/* XFS_IOC_FSBULKSTAT */
792		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
793				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
794				     &done);
795
796	if (error)
797		return error;
798
799	if (bulkreq.ocount != NULL) {
800		if (copy_to_user(bulkreq.lastip, &inlast,
801						sizeof(xfs_ino_t)))
802			return -EFAULT;
803
804		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
805			return -EFAULT;
806	}
807
808	return 0;
809}
810
811STATIC int
812xfs_ioc_fsgeometry_v1(
813	xfs_mount_t		*mp,
814	void			__user *arg)
815{
816	xfs_fsop_geom_t         fsgeo;
817	int			error;
818
819	error = xfs_fs_geometry(mp, &fsgeo, 3);
820	if (error)
821		return error;
822
823	/*
824	 * Caller should have passed an argument of type
825	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
826	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
827	 */
828	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
829		return -EFAULT;
830	return 0;
831}
832
833STATIC int
834xfs_ioc_fsgeometry(
835	xfs_mount_t		*mp,
836	void			__user *arg)
837{
838	xfs_fsop_geom_t		fsgeo;
839	int			error;
840
841	error = xfs_fs_geometry(mp, &fsgeo, 4);
842	if (error)
843		return error;
844
845	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
846		return -EFAULT;
847	return 0;
848}
849
850/*
851 * Linux extended inode flags interface.
852 */
853
854STATIC unsigned int
855xfs_merge_ioc_xflags(
856	unsigned int	flags,
857	unsigned int	start)
858{
859	unsigned int	xflags = start;
860
861	if (flags & FS_IMMUTABLE_FL)
862		xflags |= XFS_XFLAG_IMMUTABLE;
863	else
864		xflags &= ~XFS_XFLAG_IMMUTABLE;
865	if (flags & FS_APPEND_FL)
866		xflags |= XFS_XFLAG_APPEND;
867	else
868		xflags &= ~XFS_XFLAG_APPEND;
869	if (flags & FS_SYNC_FL)
870		xflags |= XFS_XFLAG_SYNC;
871	else
872		xflags &= ~XFS_XFLAG_SYNC;
873	if (flags & FS_NOATIME_FL)
874		xflags |= XFS_XFLAG_NOATIME;
875	else
876		xflags &= ~XFS_XFLAG_NOATIME;
877	if (flags & FS_NODUMP_FL)
878		xflags |= XFS_XFLAG_NODUMP;
879	else
880		xflags &= ~XFS_XFLAG_NODUMP;
881
882	return xflags;
883}
884
885STATIC unsigned int
886xfs_di2lxflags(
887	__uint16_t	di_flags)
888{
889	unsigned int	flags = 0;
890
891	if (di_flags & XFS_DIFLAG_IMMUTABLE)
892		flags |= FS_IMMUTABLE_FL;
893	if (di_flags & XFS_DIFLAG_APPEND)
894		flags |= FS_APPEND_FL;
895	if (di_flags & XFS_DIFLAG_SYNC)
896		flags |= FS_SYNC_FL;
897	if (di_flags & XFS_DIFLAG_NOATIME)
898		flags |= FS_NOATIME_FL;
899	if (di_flags & XFS_DIFLAG_NODUMP)
900		flags |= FS_NODUMP_FL;
901	return flags;
902}
903
904STATIC int
905xfs_ioc_fsgetxattr(
906	xfs_inode_t		*ip,
907	int			attr,
908	void			__user *arg)
909{
910	struct fsxattr		fa;
911
912	memset(&fa, 0, sizeof(struct fsxattr));
913
914	xfs_ilock(ip, XFS_ILOCK_SHARED);
915	fa.fsx_xflags = xfs_ip2xflags(ip);
916	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
917	fa.fsx_projid = xfs_get_projid(ip);
918
919	if (attr) {
920		if (ip->i_afp) {
921			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
922				fa.fsx_nextents = ip->i_afp->if_bytes /
923							sizeof(xfs_bmbt_rec_t);
924			else
925				fa.fsx_nextents = ip->i_d.di_anextents;
926		} else
927			fa.fsx_nextents = 0;
928	} else {
929		if (ip->i_df.if_flags & XFS_IFEXTENTS)
930			fa.fsx_nextents = ip->i_df.if_bytes /
931						sizeof(xfs_bmbt_rec_t);
932		else
933			fa.fsx_nextents = ip->i_d.di_nextents;
934	}
935	xfs_iunlock(ip, XFS_ILOCK_SHARED);
936
937	if (copy_to_user(arg, &fa, sizeof(fa)))
938		return -EFAULT;
939	return 0;
940}
941
942STATIC void
943xfs_set_diflags(
944	struct xfs_inode	*ip,
945	unsigned int		xflags)
946{
947	unsigned int		di_flags;
948
949	/* can't set PREALLOC this way, just preserve it */
950	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
951	if (xflags & XFS_XFLAG_IMMUTABLE)
952		di_flags |= XFS_DIFLAG_IMMUTABLE;
953	if (xflags & XFS_XFLAG_APPEND)
954		di_flags |= XFS_DIFLAG_APPEND;
955	if (xflags & XFS_XFLAG_SYNC)
956		di_flags |= XFS_DIFLAG_SYNC;
957	if (xflags & XFS_XFLAG_NOATIME)
958		di_flags |= XFS_DIFLAG_NOATIME;
959	if (xflags & XFS_XFLAG_NODUMP)
960		di_flags |= XFS_DIFLAG_NODUMP;
961	if (xflags & XFS_XFLAG_NODEFRAG)
962		di_flags |= XFS_DIFLAG_NODEFRAG;
963	if (xflags & XFS_XFLAG_FILESTREAM)
964		di_flags |= XFS_DIFLAG_FILESTREAM;
965	if (S_ISDIR(ip->i_d.di_mode)) {
966		if (xflags & XFS_XFLAG_RTINHERIT)
967			di_flags |= XFS_DIFLAG_RTINHERIT;
968		if (xflags & XFS_XFLAG_NOSYMLINKS)
969			di_flags |= XFS_DIFLAG_NOSYMLINKS;
970		if (xflags & XFS_XFLAG_EXTSZINHERIT)
971			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
972		if (xflags & XFS_XFLAG_PROJINHERIT)
973			di_flags |= XFS_DIFLAG_PROJINHERIT;
974	} else if (S_ISREG(ip->i_d.di_mode)) {
975		if (xflags & XFS_XFLAG_REALTIME)
976			di_flags |= XFS_DIFLAG_REALTIME;
977		if (xflags & XFS_XFLAG_EXTSIZE)
978			di_flags |= XFS_DIFLAG_EXTSIZE;
979	}
980
981	ip->i_d.di_flags = di_flags;
982}
983
984STATIC void
985xfs_diflags_to_linux(
986	struct xfs_inode	*ip)
987{
988	struct inode		*inode = VFS_I(ip);
989	unsigned int		xflags = xfs_ip2xflags(ip);
990
991	if (xflags & XFS_XFLAG_IMMUTABLE)
992		inode->i_flags |= S_IMMUTABLE;
993	else
994		inode->i_flags &= ~S_IMMUTABLE;
995	if (xflags & XFS_XFLAG_APPEND)
996		inode->i_flags |= S_APPEND;
997	else
998		inode->i_flags &= ~S_APPEND;
999	if (xflags & XFS_XFLAG_SYNC)
1000		inode->i_flags |= S_SYNC;
1001	else
1002		inode->i_flags &= ~S_SYNC;
1003	if (xflags & XFS_XFLAG_NOATIME)
1004		inode->i_flags |= S_NOATIME;
1005	else
1006		inode->i_flags &= ~S_NOATIME;
1007}
1008
1009static int
1010xfs_ioctl_setattr_xflags(
1011	struct xfs_trans	*tp,
1012	struct xfs_inode	*ip,
1013	struct fsxattr		*fa)
1014{
1015	struct xfs_mount	*mp = ip->i_mount;
1016
1017	/* Can't change realtime flag if any extents are allocated. */
1018	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1019	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & XFS_XFLAG_REALTIME))
1020		return -EINVAL;
1021
1022	/* If realtime flag is set then must have realtime device */
1023	if (fa->fsx_xflags & XFS_XFLAG_REALTIME) {
1024		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1025		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1026			return -EINVAL;
1027	}
1028
1029	/*
1030	 * Can't modify an immutable/append-only file unless
1031	 * we have appropriate permission.
1032	 */
1033	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1034	     (fa->fsx_xflags & (XFS_XFLAG_IMMUTABLE | XFS_XFLAG_APPEND))) &&
1035	    !capable(CAP_LINUX_IMMUTABLE))
1036		return -EPERM;
1037
1038	xfs_set_diflags(ip, fa->fsx_xflags);
1039	xfs_diflags_to_linux(ip);
1040	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1041	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1042	XFS_STATS_INC(mp, xs_ig_attrchg);
1043	return 0;
1044}
1045
1046/*
1047 * Set up the transaction structure for the setattr operation, checking that we
1048 * have permission to do so. On success, return a clean transaction and the
1049 * inode locked exclusively ready for further operation specific checks. On
1050 * failure, return an error without modifying or locking the inode.
1051 */
1052static struct xfs_trans *
1053xfs_ioctl_setattr_get_trans(
1054	struct xfs_inode	*ip)
1055{
1056	struct xfs_mount	*mp = ip->i_mount;
1057	struct xfs_trans	*tp;
1058	int			error;
1059
1060	if (mp->m_flags & XFS_MOUNT_RDONLY)
1061		return ERR_PTR(-EROFS);
1062	if (XFS_FORCED_SHUTDOWN(mp))
1063		return ERR_PTR(-EIO);
1064
1065	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
1066	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
1067	if (error)
1068		goto out_cancel;
1069
1070	xfs_ilock(ip, XFS_ILOCK_EXCL);
1071	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1072
1073	/*
1074	 * CAP_FOWNER overrides the following restrictions:
1075	 *
1076	 * The user ID of the calling process must be equal to the file owner
1077	 * ID, except in cases where the CAP_FSETID capability is applicable.
1078	 */
1079	if (!inode_owner_or_capable(VFS_I(ip))) {
1080		error = -EPERM;
1081		goto out_cancel;
1082	}
1083
1084	if (mp->m_flags & XFS_MOUNT_WSYNC)
1085		xfs_trans_set_sync(tp);
1086
1087	return tp;
1088
1089out_cancel:
1090	xfs_trans_cancel(tp);
1091	return ERR_PTR(error);
1092}
1093
1094/*
1095 * extent size hint validation is somewhat cumbersome. Rules are:
1096 *
1097 * 1. extent size hint is only valid for directories and regular files
1098 * 2. XFS_XFLAG_EXTSIZE is only valid for regular files
1099 * 3. XFS_XFLAG_EXTSZINHERIT is only valid for directories.
1100 * 4. can only be changed on regular files if no extents are allocated
1101 * 5. can be changed on directories at any time
1102 * 6. extsize hint of 0 turns off hints, clears inode flags.
1103 * 7. Extent size must be a multiple of the appropriate block size.
1104 * 8. for non-realtime files, the extent size hint must be limited
1105 *    to half the AG size to avoid alignment extending the extent beyond the
1106 *    limits of the AG.
1107 */
1108static int
1109xfs_ioctl_setattr_check_extsize(
1110	struct xfs_inode	*ip,
1111	struct fsxattr		*fa)
1112{
1113	struct xfs_mount	*mp = ip->i_mount;
1114
1115	if ((fa->fsx_xflags & XFS_XFLAG_EXTSIZE) && !S_ISREG(ip->i_d.di_mode))
1116		return -EINVAL;
1117
1118	if ((fa->fsx_xflags & XFS_XFLAG_EXTSZINHERIT) &&
1119	    !S_ISDIR(ip->i_d.di_mode))
1120		return -EINVAL;
1121
1122	if (S_ISREG(ip->i_d.di_mode) && ip->i_d.di_nextents &&
1123	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1124		return -EINVAL;
1125
1126	if (fa->fsx_extsize != 0) {
1127		xfs_extlen_t    size;
1128		xfs_fsblock_t   extsize_fsb;
1129
1130		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1131		if (extsize_fsb > MAXEXTLEN)
1132			return -EINVAL;
1133
1134		if (XFS_IS_REALTIME_INODE(ip) ||
1135		    (fa->fsx_xflags & XFS_XFLAG_REALTIME)) {
1136			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1137		} else {
1138			size = mp->m_sb.sb_blocksize;
1139			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1140				return -EINVAL;
1141		}
1142
1143		if (fa->fsx_extsize % size)
1144			return -EINVAL;
1145	} else
1146		fa->fsx_xflags &= ~(XFS_XFLAG_EXTSIZE | XFS_XFLAG_EXTSZINHERIT);
1147
1148	return 0;
1149}
1150
1151static int
1152xfs_ioctl_setattr_check_projid(
1153	struct xfs_inode	*ip,
1154	struct fsxattr		*fa)
1155{
1156	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1157	if (fa->fsx_projid > (__uint16_t)-1 &&
1158	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1159		return -EINVAL;
1160
1161	/*
1162	 * Project Quota ID state is only allowed to change from within the init
1163	 * namespace. Enforce that restriction only if we are trying to change
1164	 * the quota ID state. Everything else is allowed in user namespaces.
1165	 */
1166	if (current_user_ns() == &init_user_ns)
1167		return 0;
1168
1169	if (xfs_get_projid(ip) != fa->fsx_projid)
1170		return -EINVAL;
1171	if ((fa->fsx_xflags & XFS_XFLAG_PROJINHERIT) !=
1172	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1173		return -EINVAL;
1174
1175	return 0;
1176}
1177
1178STATIC int
1179xfs_ioctl_setattr(
1180	xfs_inode_t		*ip,
1181	struct fsxattr		*fa)
1182{
1183	struct xfs_mount	*mp = ip->i_mount;
1184	struct xfs_trans	*tp;
1185	struct xfs_dquot	*udqp = NULL;
1186	struct xfs_dquot	*pdqp = NULL;
1187	struct xfs_dquot	*olddquot = NULL;
1188	int			code;
1189
1190	trace_xfs_ioctl_setattr(ip);
1191
1192	code = xfs_ioctl_setattr_check_projid(ip, fa);
1193	if (code)
1194		return code;
1195
1196	/*
1197	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1198	 * before we start any other transactions. Trying to do this later
1199	 * is messy. We don't care to take a readlock to look at the ids
1200	 * in inode here, because we can't hold it across the trans_reserve.
1201	 * If the IDs do change before we take the ilock, we're covered
1202	 * because the i_*dquot fields will get updated anyway.
1203	 */
1204	if (XFS_IS_QUOTA_ON(mp)) {
1205		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1206					 ip->i_d.di_gid, fa->fsx_projid,
1207					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1208		if (code)
1209			return code;
1210	}
1211
1212	tp = xfs_ioctl_setattr_get_trans(ip);
1213	if (IS_ERR(tp)) {
1214		code = PTR_ERR(tp);
1215		goto error_free_dquots;
1216	}
1217
1218
1219	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1220	    xfs_get_projid(ip) != fa->fsx_projid) {
1221		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1222				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1223		if (code)	/* out of quota */
1224			goto error_trans_cancel;
1225	}
1226
1227	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1228	if (code)
1229		goto error_trans_cancel;
1230
1231	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1232	if (code)
1233		goto error_trans_cancel;
1234
1235	/*
1236	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1237	 * overrides the following restrictions:
1238	 *
1239	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1240	 * successful return from chown()
1241	 */
1242
1243	if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
1244	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1245		ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
1246
1247	/* Change the ownerships and register project quota modifications */
1248	if (xfs_get_projid(ip) != fa->fsx_projid) {
1249		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1250			olddquot = xfs_qm_vop_chown(tp, ip,
1251						&ip->i_pdquot, pdqp);
1252		}
1253		ASSERT(ip->i_d.di_version > 1);
1254		xfs_set_projid(ip, fa->fsx_projid);
1255	}
1256
1257	/*
1258	 * Only set the extent size hint if we've already determined that the
1259	 * extent size hint should be set on the inode. If no extent size flags
1260	 * are set on the inode then unconditionally clear the extent size hint.
1261	 */
1262	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1263		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1264	else
1265		ip->i_d.di_extsize = 0;
1266
1267	code = xfs_trans_commit(tp);
1268
1269	/*
1270	 * Release any dquot(s) the inode had kept before chown.
1271	 */
1272	xfs_qm_dqrele(olddquot);
1273	xfs_qm_dqrele(udqp);
1274	xfs_qm_dqrele(pdqp);
1275
1276	return code;
1277
1278error_trans_cancel:
1279	xfs_trans_cancel(tp);
1280error_free_dquots:
1281	xfs_qm_dqrele(udqp);
1282	xfs_qm_dqrele(pdqp);
1283	return code;
1284}
1285
1286STATIC int
1287xfs_ioc_fssetxattr(
1288	xfs_inode_t		*ip,
1289	struct file		*filp,
1290	void			__user *arg)
1291{
1292	struct fsxattr		fa;
1293	int error;
1294
1295	if (copy_from_user(&fa, arg, sizeof(fa)))
1296		return -EFAULT;
1297
1298	error = mnt_want_write_file(filp);
1299	if (error)
1300		return error;
1301	error = xfs_ioctl_setattr(ip, &fa);
1302	mnt_drop_write_file(filp);
1303	return error;
1304}
1305
1306STATIC int
1307xfs_ioc_getxflags(
1308	xfs_inode_t		*ip,
1309	void			__user *arg)
1310{
1311	unsigned int		flags;
1312
1313	flags = xfs_di2lxflags(ip->i_d.di_flags);
1314	if (copy_to_user(arg, &flags, sizeof(flags)))
1315		return -EFAULT;
1316	return 0;
1317}
1318
1319STATIC int
1320xfs_ioc_setxflags(
1321	struct xfs_inode	*ip,
1322	struct file		*filp,
1323	void			__user *arg)
1324{
1325	struct xfs_trans	*tp;
1326	struct fsxattr		fa;
1327	unsigned int		flags;
1328	int			error;
1329
1330	if (copy_from_user(&flags, arg, sizeof(flags)))
1331		return -EFAULT;
1332
1333	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1334		      FS_NOATIME_FL | FS_NODUMP_FL | \
1335		      FS_SYNC_FL))
1336		return -EOPNOTSUPP;
1337
1338	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1339
1340	error = mnt_want_write_file(filp);
1341	if (error)
1342		return error;
1343
1344	tp = xfs_ioctl_setattr_get_trans(ip);
1345	if (IS_ERR(tp)) {
1346		error = PTR_ERR(tp);
1347		goto out_drop_write;
1348	}
1349
1350	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1351	if (error) {
1352		xfs_trans_cancel(tp);
1353		goto out_drop_write;
1354	}
1355
1356	error = xfs_trans_commit(tp);
1357out_drop_write:
1358	mnt_drop_write_file(filp);
1359	return error;
1360}
1361
1362STATIC int
1363xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
1364{
1365	struct getbmap __user	*base = (struct getbmap __user *)*ap;
1366
1367	/* copy only getbmap portion (not getbmapx) */
1368	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1369		return -EFAULT;
1370
1371	*ap += sizeof(struct getbmap);
1372	return 0;
1373}
1374
1375STATIC int
1376xfs_ioc_getbmap(
1377	struct xfs_inode	*ip,
1378	int			ioflags,
1379	unsigned int		cmd,
1380	void			__user *arg)
1381{
1382	struct getbmapx		bmx;
1383	int			error;
1384
1385	if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
1386		return -EFAULT;
1387
1388	if (bmx.bmv_count < 2)
1389		return -EINVAL;
1390
1391	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1392	if (ioflags & XFS_IO_INVIS)
1393		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1394
1395	error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
1396			    (__force struct getbmap *)arg+1);
1397	if (error)
1398		return error;
1399
1400	/* copy back header - only size of getbmap */
1401	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1402		return -EFAULT;
1403	return 0;
1404}
1405
1406STATIC int
1407xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
1408{
1409	struct getbmapx __user	*base = (struct getbmapx __user *)*ap;
1410
1411	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1412		return -EFAULT;
1413
1414	*ap += sizeof(struct getbmapx);
1415	return 0;
1416}
1417
1418STATIC int
1419xfs_ioc_getbmapx(
1420	struct xfs_inode	*ip,
1421	void			__user *arg)
1422{
1423	struct getbmapx		bmx;
1424	int			error;
1425
1426	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1427		return -EFAULT;
1428
1429	if (bmx.bmv_count < 2)
1430		return -EINVAL;
1431
1432	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1433		return -EINVAL;
1434
1435	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1436			    (__force struct getbmapx *)arg+1);
1437	if (error)
1438		return error;
1439
1440	/* copy back header */
1441	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1442		return -EFAULT;
1443
1444	return 0;
1445}
1446
1447int
1448xfs_ioc_swapext(
1449	xfs_swapext_t	*sxp)
1450{
1451	xfs_inode_t     *ip, *tip;
1452	struct fd	f, tmp;
1453	int		error = 0;
1454
1455	/* Pull information for the target fd */
1456	f = fdget((int)sxp->sx_fdtarget);
1457	if (!f.file) {
1458		error = -EINVAL;
1459		goto out;
1460	}
1461
1462	if (!(f.file->f_mode & FMODE_WRITE) ||
1463	    !(f.file->f_mode & FMODE_READ) ||
1464	    (f.file->f_flags & O_APPEND)) {
1465		error = -EBADF;
1466		goto out_put_file;
1467	}
1468
1469	tmp = fdget((int)sxp->sx_fdtmp);
1470	if (!tmp.file) {
1471		error = -EINVAL;
1472		goto out_put_file;
1473	}
1474
1475	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1476	    !(tmp.file->f_mode & FMODE_READ) ||
1477	    (tmp.file->f_flags & O_APPEND)) {
1478		error = -EBADF;
1479		goto out_put_tmp_file;
1480	}
1481
1482	if (IS_SWAPFILE(file_inode(f.file)) ||
1483	    IS_SWAPFILE(file_inode(tmp.file))) {
1484		error = -EINVAL;
1485		goto out_put_tmp_file;
1486	}
1487
1488	ip = XFS_I(file_inode(f.file));
1489	tip = XFS_I(file_inode(tmp.file));
1490
1491	if (ip->i_mount != tip->i_mount) {
1492		error = -EINVAL;
1493		goto out_put_tmp_file;
1494	}
1495
1496	if (ip->i_ino == tip->i_ino) {
1497		error = -EINVAL;
1498		goto out_put_tmp_file;
1499	}
1500
1501	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1502		error = -EIO;
1503		goto out_put_tmp_file;
1504	}
1505
1506	error = xfs_swap_extents(ip, tip, sxp);
1507
1508 out_put_tmp_file:
1509	fdput(tmp);
1510 out_put_file:
1511	fdput(f);
1512 out:
1513	return error;
1514}
1515
1516/*
1517 * Note: some of the ioctl's return positive numbers as a
1518 * byte count indicating success, such as readlink_by_handle.
1519 * So we don't "sign flip" like most other routines.  This means
1520 * true errors need to be returned as a negative value.
1521 */
1522long
1523xfs_file_ioctl(
1524	struct file		*filp,
1525	unsigned int		cmd,
1526	unsigned long		p)
1527{
1528	struct inode		*inode = file_inode(filp);
1529	struct xfs_inode	*ip = XFS_I(inode);
1530	struct xfs_mount	*mp = ip->i_mount;
1531	void			__user *arg = (void __user *)p;
1532	int			ioflags = 0;
1533	int			error;
1534
1535	if (filp->f_mode & FMODE_NOCMTIME)
1536		ioflags |= XFS_IO_INVIS;
1537
1538	trace_xfs_file_ioctl(ip);
1539
1540	switch (cmd) {
1541	case FITRIM:
1542		return xfs_ioc_trim(mp, arg);
1543	case XFS_IOC_ALLOCSP:
1544	case XFS_IOC_FREESP:
1545	case XFS_IOC_RESVSP:
1546	case XFS_IOC_UNRESVSP:
1547	case XFS_IOC_ALLOCSP64:
1548	case XFS_IOC_FREESP64:
1549	case XFS_IOC_RESVSP64:
1550	case XFS_IOC_UNRESVSP64:
1551	case XFS_IOC_ZERO_RANGE: {
1552		xfs_flock64_t		bf;
1553
1554		if (copy_from_user(&bf, arg, sizeof(bf)))
1555			return -EFAULT;
1556		return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
1557	}
1558	case XFS_IOC_DIOINFO: {
1559		struct dioattr	da;
1560		xfs_buftarg_t	*target =
1561			XFS_IS_REALTIME_INODE(ip) ?
1562			mp->m_rtdev_targp : mp->m_ddev_targp;
1563
1564		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1565		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1566
1567		if (copy_to_user(arg, &da, sizeof(da)))
1568			return -EFAULT;
1569		return 0;
1570	}
1571
1572	case XFS_IOC_FSBULKSTAT_SINGLE:
1573	case XFS_IOC_FSBULKSTAT:
1574	case XFS_IOC_FSINUMBERS:
1575		return xfs_ioc_bulkstat(mp, cmd, arg);
1576
1577	case XFS_IOC_FSGEOMETRY_V1:
1578		return xfs_ioc_fsgeometry_v1(mp, arg);
1579
1580	case XFS_IOC_FSGEOMETRY:
1581		return xfs_ioc_fsgeometry(mp, arg);
1582
1583	case XFS_IOC_GETVERSION:
1584		return put_user(inode->i_generation, (int __user *)arg);
1585
1586	case XFS_IOC_FSGETXATTR:
1587		return xfs_ioc_fsgetxattr(ip, 0, arg);
1588	case XFS_IOC_FSGETXATTRA:
1589		return xfs_ioc_fsgetxattr(ip, 1, arg);
1590	case XFS_IOC_FSSETXATTR:
1591		return xfs_ioc_fssetxattr(ip, filp, arg);
1592	case XFS_IOC_GETXFLAGS:
1593		return xfs_ioc_getxflags(ip, arg);
1594	case XFS_IOC_SETXFLAGS:
1595		return xfs_ioc_setxflags(ip, filp, arg);
1596
1597	case XFS_IOC_FSSETDM: {
1598		struct fsdmidata	dmi;
1599
1600		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1601			return -EFAULT;
1602
1603		error = mnt_want_write_file(filp);
1604		if (error)
1605			return error;
1606
1607		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1608				dmi.fsd_dmstate);
1609		mnt_drop_write_file(filp);
1610		return error;
1611	}
1612
1613	case XFS_IOC_GETBMAP:
1614	case XFS_IOC_GETBMAPA:
1615		return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
1616
1617	case XFS_IOC_GETBMAPX:
1618		return xfs_ioc_getbmapx(ip, arg);
1619
1620	case XFS_IOC_FD_TO_HANDLE:
1621	case XFS_IOC_PATH_TO_HANDLE:
1622	case XFS_IOC_PATH_TO_FSHANDLE: {
1623		xfs_fsop_handlereq_t	hreq;
1624
1625		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1626			return -EFAULT;
1627		return xfs_find_handle(cmd, &hreq);
1628	}
1629	case XFS_IOC_OPEN_BY_HANDLE: {
1630		xfs_fsop_handlereq_t	hreq;
1631
1632		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1633			return -EFAULT;
1634		return xfs_open_by_handle(filp, &hreq);
1635	}
1636	case XFS_IOC_FSSETDM_BY_HANDLE:
1637		return xfs_fssetdm_by_handle(filp, arg);
1638
1639	case XFS_IOC_READLINK_BY_HANDLE: {
1640		xfs_fsop_handlereq_t	hreq;
1641
1642		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1643			return -EFAULT;
1644		return xfs_readlink_by_handle(filp, &hreq);
1645	}
1646	case XFS_IOC_ATTRLIST_BY_HANDLE:
1647		return xfs_attrlist_by_handle(filp, arg);
1648
1649	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1650		return xfs_attrmulti_by_handle(filp, arg);
1651
1652	case XFS_IOC_SWAPEXT: {
1653		struct xfs_swapext	sxp;
1654
1655		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1656			return -EFAULT;
1657		error = mnt_want_write_file(filp);
1658		if (error)
1659			return error;
1660		error = xfs_ioc_swapext(&sxp);
1661		mnt_drop_write_file(filp);
1662		return error;
1663	}
1664
1665	case XFS_IOC_FSCOUNTS: {
1666		xfs_fsop_counts_t out;
1667
1668		error = xfs_fs_counts(mp, &out);
1669		if (error)
1670			return error;
1671
1672		if (copy_to_user(arg, &out, sizeof(out)))
1673			return -EFAULT;
1674		return 0;
1675	}
1676
1677	case XFS_IOC_SET_RESBLKS: {
1678		xfs_fsop_resblks_t inout;
1679		__uint64_t	   in;
1680
1681		if (!capable(CAP_SYS_ADMIN))
1682			return -EPERM;
1683
1684		if (mp->m_flags & XFS_MOUNT_RDONLY)
1685			return -EROFS;
1686
1687		if (copy_from_user(&inout, arg, sizeof(inout)))
1688			return -EFAULT;
1689
1690		error = mnt_want_write_file(filp);
1691		if (error)
1692			return error;
1693
1694		/* input parameter is passed in resblks field of structure */
1695		in = inout.resblks;
1696		error = xfs_reserve_blocks(mp, &in, &inout);
1697		mnt_drop_write_file(filp);
1698		if (error)
1699			return error;
1700
1701		if (copy_to_user(arg, &inout, sizeof(inout)))
1702			return -EFAULT;
1703		return 0;
1704	}
1705
1706	case XFS_IOC_GET_RESBLKS: {
1707		xfs_fsop_resblks_t out;
1708
1709		if (!capable(CAP_SYS_ADMIN))
1710			return -EPERM;
1711
1712		error = xfs_reserve_blocks(mp, NULL, &out);
1713		if (error)
1714			return error;
1715
1716		if (copy_to_user(arg, &out, sizeof(out)))
1717			return -EFAULT;
1718
1719		return 0;
1720	}
1721
1722	case XFS_IOC_FSGROWFSDATA: {
1723		xfs_growfs_data_t in;
1724
1725		if (copy_from_user(&in, arg, sizeof(in)))
1726			return -EFAULT;
1727
1728		error = mnt_want_write_file(filp);
1729		if (error)
1730			return error;
1731		error = xfs_growfs_data(mp, &in);
1732		mnt_drop_write_file(filp);
1733		return error;
1734	}
1735
1736	case XFS_IOC_FSGROWFSLOG: {
1737		xfs_growfs_log_t in;
1738
1739		if (copy_from_user(&in, arg, sizeof(in)))
1740			return -EFAULT;
1741
1742		error = mnt_want_write_file(filp);
1743		if (error)
1744			return error;
1745		error = xfs_growfs_log(mp, &in);
1746		mnt_drop_write_file(filp);
1747		return error;
1748	}
1749
1750	case XFS_IOC_FSGROWFSRT: {
1751		xfs_growfs_rt_t in;
1752
1753		if (copy_from_user(&in, arg, sizeof(in)))
1754			return -EFAULT;
1755
1756		error = mnt_want_write_file(filp);
1757		if (error)
1758			return error;
1759		error = xfs_growfs_rt(mp, &in);
1760		mnt_drop_write_file(filp);
1761		return error;
1762	}
1763
1764	case XFS_IOC_GOINGDOWN: {
1765		__uint32_t in;
1766
1767		if (!capable(CAP_SYS_ADMIN))
1768			return -EPERM;
1769
1770		if (get_user(in, (__uint32_t __user *)arg))
1771			return -EFAULT;
1772
1773		return xfs_fs_goingdown(mp, in);
1774	}
1775
1776	case XFS_IOC_ERROR_INJECTION: {
1777		xfs_error_injection_t in;
1778
1779		if (!capable(CAP_SYS_ADMIN))
1780			return -EPERM;
1781
1782		if (copy_from_user(&in, arg, sizeof(in)))
1783			return -EFAULT;
1784
1785		return xfs_errortag_add(in.errtag, mp);
1786	}
1787
1788	case XFS_IOC_ERROR_CLEARALL:
1789		if (!capable(CAP_SYS_ADMIN))
1790			return -EPERM;
1791
1792		return xfs_errortag_clearall(mp, 1);
1793
1794	case XFS_IOC_FREE_EOFBLOCKS: {
1795		struct xfs_fs_eofblocks eofb;
1796		struct xfs_eofblocks keofb;
1797
1798		if (!capable(CAP_SYS_ADMIN))
1799			return -EPERM;
1800
1801		if (mp->m_flags & XFS_MOUNT_RDONLY)
1802			return -EROFS;
1803
1804		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1805			return -EFAULT;
1806
1807		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
1808		if (error)
1809			return error;
1810
1811		return xfs_icache_free_eofblocks(mp, &keofb);
1812	}
1813
1814	default:
1815		return -ENOTTY;
1816	}
1817}
1818