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