1/*
2 * Copyright (c) 2000-2006 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_types.h"
21#include "xfs_bit.h"
22#include "xfs_log.h"
23#include "xfs_inum.h"
24#include "xfs_trans.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_dir.h"
28#include "xfs_dir2.h"
29#include "xfs_dmapi.h"
30#include "xfs_mount.h"
31#include "xfs_bmap_btree.h"
32#include "xfs_alloc_btree.h"
33#include "xfs_ialloc_btree.h"
34#include "xfs_dir_sf.h"
35#include "xfs_dir2_sf.h"
36#include "xfs_attr_sf.h"
37#include "xfs_dinode.h"
38#include "xfs_inode.h"
39#include "xfs_inode_item.h"
40#include "xfs_bmap.h"
41#include "xfs_btree.h"
42#include "xfs_ialloc.h"
43#include "xfs_itable.h"
44#include "xfs_dfrag.h"
45#include "xfs_error.h"
46#include "xfs_mac.h"
47#include "xfs_rw.h"
48
49#include <sys/capability.h>
50#include <sys/file.h>
51
52/*
53 * Syssgi interface for swapext
54 */
55int
56xfs_swapext(
57	xfs_swapext_t	__user *sxu)
58{
59	xfs_swapext_t	*sxp;
60	xfs_inode_t     *ip=NULL, *tip=NULL;
61	xfs_mount_t     *mp;
62	xfs_vnode_t	*vp = NULL, *tvp = NULL;
63	struct vnode	*bvp, *btvp;
64	int		error = 0;
65
66	sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
67	if (!sxp) {
68		error = XFS_ERROR(ENOMEM);
69		goto error0;
70	}
71	struct thread	*td;
72	struct cred	*cred;
73
74	td = curthread;
75	cred = td->td_ucred;
76
77	if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
78		error = XFS_ERROR(EFAULT);
79		goto error0;
80	}
81
82	/* Pull information for the target fd */
83	if (fgetvp(td, (int)sxp->sx_fdtarget, CAP_READ | CAP_WRITE, &bvp)
84	    != 0) {
85		error = XFS_ERROR(EINVAL);
86		goto error0;
87	}
88
89	vp = VPTOXFSVP(bvp);
90	ip = xfs_vtoi(vp);
91	if (ip == NULL) {
92		error = XFS_ERROR(EBADF);
93		goto error0;
94	}
95
96	if (fgetvp(td, (int)sxp->sx_fdtmp, CAP_READ | CAP_WRITE, &btvp) != 0) {
97		error = XFS_ERROR(EINVAL);
98		goto error0;
99	}
100
101	tvp = VPTOXFSVP(btvp);
102	tip = xfs_vtoi(tvp);
103	if (tip == NULL) {
104		error = XFS_ERROR(EBADF);
105		goto error0;
106	}
107
108	if (ip->i_mount != tip->i_mount) {
109		error =  XFS_ERROR(EINVAL);
110		goto error0;
111	}
112
113	if (ip->i_ino == tip->i_ino) {
114		error =  XFS_ERROR(EINVAL);
115		goto error0;
116	}
117
118	mp = ip->i_mount;
119
120	if (XFS_FORCED_SHUTDOWN(mp)) {
121		error =  XFS_ERROR(EIO);
122		goto error0;
123	}
124
125	error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
126
127 error0:
128#ifdef RMC
129	if (fp != NULL)
130		fput(fp);
131	if (tfp != NULL)
132		fput(tfp);
133#endif
134
135	if (sxp != NULL)
136		kmem_free(sxp, sizeof(xfs_swapext_t));
137
138	return error;
139}
140
141int
142xfs_swap_extents(
143	xfs_inode_t	*ip,
144	xfs_inode_t	*tip,
145	xfs_swapext_t	*sxp)
146{
147	xfs_mount_t	*mp;
148	xfs_inode_t	*ips[2];
149	xfs_trans_t	*tp;
150	xfs_bstat_t	*sbp = &sxp->sx_stat;
151	xfs_vnode_t	*vp, *tvp;
152	xfs_ifork_t	*tempifp, *ifp, *tifp;
153	int		ilf_fields, tilf_fields;
154	static uint	lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
155	int		error = 0;
156	int		aforkblks = 0;
157	int		taforkblks = 0;
158	__uint64_t	tmp;
159	char		locked = 0;
160
161	mp = ip->i_mount;
162
163	tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
164	if (!tempifp) {
165		error = XFS_ERROR(ENOMEM);
166		goto error0;
167	}
168
169	sbp = &sxp->sx_stat;
170	vp = XFS_ITOV(ip);
171	tvp = XFS_ITOV(tip);
172
173	/* Lock in i_ino order */
174	if (ip->i_ino < tip->i_ino) {
175		ips[0] = ip;
176		ips[1] = tip;
177	} else {
178		ips[0] = tip;
179		ips[1] = ip;
180	}
181
182	xfs_lock_inodes(ips, 2, 0, lock_flags);
183	locked = 1;
184
185	/* Check permissions */
186	error = xfs_iaccess(ip, VWRITE, NULL);
187	if (error)
188		goto error0;
189
190	error = xfs_iaccess(tip, VWRITE, NULL);
191	if (error)
192		goto error0;
193
194	/* Verify that both files have the same format */
195	if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
196		error = XFS_ERROR(EINVAL);
197		goto error0;
198	}
199
200	/* Verify both files are either real-time or non-realtime */
201	if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
202	    (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
203		error = XFS_ERROR(EINVAL);
204		goto error0;
205	}
206
207	/* Should never get a local format */
208	if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
209	    tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
210		error = XFS_ERROR(EINVAL);
211		goto error0;
212	}
213
214	if (VN_CACHED(tvp) != 0) {
215		xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
216		XVOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
217	}
218
219	/* Verify O_DIRECT for ftmp */
220	if (VN_CACHED(tvp) != 0) {
221		error = XFS_ERROR(EINVAL);
222		goto error0;
223	}
224
225	/* Verify all data are being swapped */
226	if (sxp->sx_offset != 0 ||
227	    sxp->sx_length != ip->i_d.di_size ||
228	    sxp->sx_length != tip->i_d.di_size) {
229		error = XFS_ERROR(EFAULT);
230		goto error0;
231	}
232
233	/*
234	 * If the target has extended attributes, the tmp file
235	 * must also in order to ensure the correct data fork
236	 * format.
237	 */
238	if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
239		error = XFS_ERROR(EINVAL);
240		goto error0;
241	}
242
243	/*
244	 * Compare the current change & modify times with that
245	 * passed in.  If they differ, we abort this swap.
246	 * This is the mechanism used to ensure the calling
247	 * process that the file was not changed out from
248	 * under it.
249	 */
250	if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
251	    (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
252	    (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
253	    (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
254		error = XFS_ERROR(EBUSY);
255		goto error0;
256	}
257
258	/* We need to fail if the file is memory mapped.  Once we have tossed
259	 * all existing pages, the page fault will have no option
260	 * but to go to the filesystem for pages. By making the page fault call
261	 * VOP_READ (or write in the case of autogrow) they block on the iolock
262	 * until we have switched the extents.
263	 */
264	if (VN_MAPPED(vp)) {
265		error = XFS_ERROR(EBUSY);
266		goto error0;
267	}
268
269	xfs_iunlock(ip, XFS_ILOCK_EXCL);
270	xfs_iunlock(tip, XFS_ILOCK_EXCL);
271
272	/*
273	 * There is a race condition here since we gave up the
274	 * ilock.  However, the data fork will not change since
275	 * we have the iolock (locked for truncation too) so we
276	 * are safe.  We don't really care if non-io related
277	 * fields change.
278	 */
279
280	XVOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
281
282	tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
283	if ((error = xfs_trans_reserve(tp, 0,
284				     XFS_ICHANGE_LOG_RES(mp), 0,
285				     0, 0))) {
286		xfs_iunlock(ip,  XFS_IOLOCK_EXCL);
287		xfs_iunlock(tip, XFS_IOLOCK_EXCL);
288		xfs_trans_cancel(tp, 0);
289		locked = 0;
290		goto error0;
291	}
292	xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
293
294	/*
295	 * Count the number of extended attribute blocks
296	 */
297	if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
298	     (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
299		error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
300		if (error) {
301			xfs_trans_cancel(tp, 0);
302			goto error0;
303		}
304	}
305	if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
306	     (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
307		error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
308			&taforkblks);
309		if (error) {
310			xfs_trans_cancel(tp, 0);
311			goto error0;
312		}
313	}
314
315	/*
316	 * Swap the data forks of the inodes
317	 */
318	ifp = &ip->i_df;
319	tifp = &tip->i_df;
320	*tempifp = *ifp;	/* struct copy */
321	*ifp = *tifp;		/* struct copy */
322	*tifp = *tempifp;	/* struct copy */
323
324	/*
325	 * Fix the on-disk inode values
326	 */
327	tmp = (__uint64_t)ip->i_d.di_nblocks;
328	ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
329	tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
330
331	tmp = (__uint64_t) ip->i_d.di_nextents;
332	ip->i_d.di_nextents = tip->i_d.di_nextents;
333	tip->i_d.di_nextents = tmp;
334
335	tmp = (__uint64_t) ip->i_d.di_format;
336	ip->i_d.di_format = tip->i_d.di_format;
337	tip->i_d.di_format = tmp;
338
339	ilf_fields = XFS_ILOG_CORE;
340
341	switch(ip->i_d.di_format) {
342	case XFS_DINODE_FMT_EXTENTS:
343		/* If the extents fit in the inode, fix the
344		 * pointer.  Otherwise it's already NULL or
345		 * pointing to the extent.
346		 */
347		if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
348			ifp->if_u1.if_extents =
349				ifp->if_u2.if_inline_ext;
350		}
351		ilf_fields |= XFS_ILOG_DEXT;
352		break;
353	case XFS_DINODE_FMT_BTREE:
354		ilf_fields |= XFS_ILOG_DBROOT;
355		break;
356	}
357
358	tilf_fields = XFS_ILOG_CORE;
359
360	switch(tip->i_d.di_format) {
361	case XFS_DINODE_FMT_EXTENTS:
362		/* If the extents fit in the inode, fix the
363		 * pointer.  Otherwise it's already NULL or
364		 * pointing to the extent.
365		 */
366		if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
367			tifp->if_u1.if_extents =
368				tifp->if_u2.if_inline_ext;
369		}
370		tilf_fields |= XFS_ILOG_DEXT;
371		break;
372	case XFS_DINODE_FMT_BTREE:
373		tilf_fields |= XFS_ILOG_DBROOT;
374		break;
375	}
376
377#ifdef XXXKAN /* Not necessary, vnodes are vrefed already by fgetvp */
378	/*
379	 * Increment vnode ref counts since xfs_trans_commit &
380	 * xfs_trans_cancel will both unlock the inodes and
381	 * decrement the associated ref counts.
382	 */
383	VN_HOLD(vp);
384	VN_HOLD(tvp);
385#endif
386
387	xfs_trans_ijoin(tp, ip, lock_flags);
388	xfs_trans_ijoin(tp, tip, lock_flags);
389
390	xfs_trans_log_inode(tp, ip,  ilf_fields);
391	xfs_trans_log_inode(tp, tip, tilf_fields);
392
393	/*
394	 * If this is a synchronous mount, make sure that the
395	 * transaction goes to disk before returning to the user.
396	 */
397	if (mp->m_flags & XFS_MOUNT_WSYNC) {
398		xfs_trans_set_sync(tp);
399	}
400
401	error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
402
403	locked = 0;
404
405 error0:
406	if (locked) {
407		xfs_iunlock(ip,  lock_flags);
408		xfs_iunlock(tip, lock_flags);
409	}
410	if (tempifp != NULL)
411		kmem_free(tempifp, sizeof(xfs_ifork_t));
412	return error;
413}
414