1/*-
2 * Copyright (c) 1989, 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	@(#)ufs_bmap.c	8.7 (Berkeley) 3/21/95
35 * $FreeBSD: stable/11/sys/fs/ext2fs/ext2_bmap.c 311231 2017-01-04 02:42:17Z pfg $
36 */
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/bio.h>
41#include <sys/buf.h>
42#include <sys/proc.h>
43#include <sys/vnode.h>
44#include <sys/mount.h>
45#include <sys/racct.h>
46#include <sys/resourcevar.h>
47#include <sys/stat.h>
48
49#include <fs/ext2fs/inode.h>
50#include <fs/ext2fs/fs.h>
51#include <fs/ext2fs/ext2fs.h>
52#include <fs/ext2fs/ext2_dinode.h>
53#include <fs/ext2fs/ext2_extern.h>
54#include <fs/ext2fs/ext2_mount.h>
55
56static int ext4_bmapext(struct vnode *, int32_t, int64_t *, int *, int *);
57
58/*
59 * Bmap converts the logical block number of a file to its physical block
60 * number on the disk. The conversion is done by using the logical block
61 * number to index into the array of block pointers described by the dinode.
62 */
63int
64ext2_bmap(struct vop_bmap_args *ap)
65{
66	daddr_t blkno;
67	int error;
68
69	/*
70	 * Check for underlying vnode requests and ensure that logical
71	 * to physical mapping is requested.
72	 */
73	if (ap->a_bop != NULL)
74		*ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj;
75	if (ap->a_bnp == NULL)
76		return (0);
77
78	if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS)
79		error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno,
80		    ap->a_runp, ap->a_runb);
81	else
82		error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno,
83		    ap->a_runp, ap->a_runb);
84	*ap->a_bnp = blkno;
85	return (error);
86}
87
88/*
89 * Convert the logical block number of a file to its physical block number
90 * on the disk within ext4 extents.
91 */
92static int
93ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb)
94{
95	struct inode *ip;
96	struct m_ext2fs *fs;
97	struct ext4_extent *ep;
98	struct ext4_extent_path path = {.ep_bp = NULL};
99	daddr_t lbn;
100	int error;
101
102	ip = VTOI(vp);
103	fs = ip->i_e2fs;
104	lbn = bn;
105
106	if (runp != NULL)
107		*runp = 0;
108	if (runb != NULL)
109		*runb = 0;
110	error = 0;
111
112	ext4_ext_find_extent(fs, ip, lbn, &path);
113	if (path.ep_is_sparse) {
114		*bnp = -1;
115		if (runp != NULL)
116			*runp = path.ep_sparse_ext.e_len -
117			    (lbn - path.ep_sparse_ext.e_blk) - 1;
118		if (runb != NULL)
119			*runb = lbn - path.ep_sparse_ext.e_blk;
120	} else {
121		if (path.ep_ext == NULL) {
122			error = EIO;
123			goto out;
124		}
125		ep = path.ep_ext;
126		*bnp = fsbtodb(fs, lbn - ep->e_blk +
127		    (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32));
128
129		if (*bnp == 0)
130			*bnp = -1;
131
132		if (runp != NULL)
133			*runp = ep->e_len - (lbn - ep->e_blk) - 1;
134		if (runb != NULL)
135			*runb = lbn - ep->e_blk;
136	}
137
138out:
139	if (path.ep_bp != NULL)
140		brelse(path.ep_bp);
141
142	return (error);
143}
144
145/*
146 * Indirect blocks are now on the vnode for the file.  They are given negative
147 * logical block numbers.  Indirect blocks are addressed by the negative
148 * address of the first data block to which they point.  Double indirect blocks
149 * are addressed by one less than the address of the first indirect block to
150 * which they point.  Triple indirect blocks are addressed by one less than
151 * the address of the first double indirect block to which they point.
152 *
153 * ext2_bmaparray does the bmap conversion, and if requested returns the
154 * array of logical blocks which must be traversed to get to a block.
155 * Each entry contains the offset into that block that gets you to the
156 * next block and the disk address of the block (if it is assigned).
157 */
158
159int
160ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
161{
162	struct inode *ip;
163	struct buf *bp;
164	struct ext2mount *ump;
165	struct mount *mp;
166	struct indir a[NIADDR + 1], *ap;
167	daddr_t daddr;
168	e2fs_lbn_t metalbn;
169	int error, num, maxrun = 0, bsize;
170	int *nump;
171
172	ap = NULL;
173	ip = VTOI(vp);
174	mp = vp->v_mount;
175	ump = VFSTOEXT2(mp);
176
177	bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
178
179	if (runp) {
180		maxrun = mp->mnt_iosize_max / bsize - 1;
181		*runp = 0;
182	}
183	if (runb)
184		*runb = 0;
185
186
187	ap = a;
188	nump = &num;
189	error = ext2_getlbns(vp, bn, ap, nump);
190	if (error)
191		return (error);
192
193	num = *nump;
194	if (num == 0) {
195		*bnp = blkptrtodb(ump, ip->i_db[bn]);
196		if (*bnp == 0) {
197			*bnp = -1;
198		} else if (runp) {
199			daddr_t bnb = bn;
200
201			for (++bn; bn < NDADDR && *runp < maxrun &&
202			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
203			    ++bn, ++*runp);
204			bn = bnb;
205			if (runb && (bn > 0)) {
206				for (--bn; (bn >= 0) && (*runb < maxrun) &&
207					is_sequential(ump, ip->i_db[bn],
208						ip->i_db[bn + 1]);
209						--bn, ++*runb);
210			}
211		}
212		return (0);
213	}
214
215	/* Get disk address out of indirect block array */
216	daddr = ip->i_ib[ap->in_off];
217
218	for (bp = NULL, ++ap; --num; ++ap) {
219		/*
220		 * Exit the loop if there is no disk address assigned yet and
221		 * the indirect block isn't in the cache, or if we were
222		 * looking for an indirect block and we've found it.
223		 */
224
225		metalbn = ap->in_lbn;
226		if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
227			break;
228		/*
229		 * If we get here, we've either got the block in the cache
230		 * or we have a disk address for it, go fetch it.
231		 */
232		if (bp)
233			bqrelse(bp);
234
235		bp = getblk(vp, metalbn, bsize, 0, 0, 0);
236		if ((bp->b_flags & B_CACHE) == 0) {
237#ifdef INVARIANTS
238			if (!daddr)
239				panic("ext2_bmaparray: indirect block not in cache");
240#endif
241			bp->b_blkno = blkptrtodb(ump, daddr);
242			bp->b_iocmd = BIO_READ;
243			bp->b_flags &= ~B_INVAL;
244			bp->b_ioflags &= ~BIO_ERROR;
245			vfs_busy_pages(bp, 0);
246			bp->b_iooffset = dbtob(bp->b_blkno);
247			bstrategy(bp);
248#ifdef RACCT
249			if (racct_enable) {
250				PROC_LOCK(curproc);
251				racct_add_buf(curproc, bp, 0);
252				PROC_UNLOCK(curproc);
253			}
254#endif
255			curthread->td_ru.ru_inblock++;
256			error = bufwait(bp);
257			if (error) {
258				brelse(bp);
259				return (error);
260			}
261		}
262
263		daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
264		if (num == 1 && daddr && runp) {
265			for (bn = ap->in_off + 1;
266			    bn < MNINDIR(ump) && *runp < maxrun &&
267			    is_sequential(ump,
268			    ((e2fs_daddr_t *)bp->b_data)[bn - 1],
269			    ((e2fs_daddr_t *)bp->b_data)[bn]);
270			    ++bn, ++*runp);
271			bn = ap->in_off;
272			if (runb && bn) {
273				for (--bn; bn >= 0 && *runb < maxrun &&
274					is_sequential(ump,
275					((e2fs_daddr_t *)bp->b_data)[bn],
276					((e2fs_daddr_t *)bp->b_data)[bn + 1]);
277					--bn, ++*runb);
278			}
279		}
280	}
281	if (bp)
282		bqrelse(bp);
283
284	/*
285	 * Since this is FFS independent code, we are out of scope for the
286	 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
287	 * will fall in the range 1..um_seqinc, so we use that test and
288	 * return a request for a zeroed out buffer if attempts are made
289	 * to read a BLK_NOCOPY or BLK_SNAP block.
290	 */
291	if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc) {
292		*bnp = -1;
293		return (0);
294	}
295	*bnp = blkptrtodb(ump, daddr);
296	if (*bnp == 0) {
297		*bnp = -1;
298	}
299	return (0);
300}
301
302/*
303 * Create an array of logical block number/offset pairs which represent the
304 * path of indirect blocks required to access a data block.  The first "pair"
305 * contains the logical block number of the appropriate single, double or
306 * triple indirect block and the offset into the inode indirect block array.
307 * Note, the logical block number of the inode single/double/triple indirect
308 * block appears twice in the array, once with the offset into the i_ib and
309 * once with the offset into the page itself.
310 */
311int
312ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
313{
314	long blockcnt;
315	e2fs_lbn_t metalbn, realbn;
316	struct ext2mount *ump;
317	int i, numlevels, off;
318	int64_t qblockcnt;
319
320	ump = VFSTOEXT2(vp->v_mount);
321	if (nump)
322		*nump = 0;
323	numlevels = 0;
324	realbn = bn;
325	if ((long)bn < 0)
326		bn = -(long)bn;
327
328	/* The first NDADDR blocks are direct blocks. */
329	if (bn < NDADDR)
330		return (0);
331
332	/*
333	 * Determine the number of levels of indirection.  After this loop
334	 * is done, blockcnt indicates the number of data blocks possible
335	 * at the previous level of indirection, and NIADDR - i is the number
336	 * of levels of indirection needed to locate the requested block.
337	 */
338	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
339		if (i == 0)
340			return (EFBIG);
341		/*
342		 * Use int64_t's here to avoid overflow for triple indirect
343		 * blocks when longs have 32 bits and the block size is more
344		 * than 4K.
345		 */
346		qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
347		if (bn < qblockcnt)
348			break;
349		blockcnt = qblockcnt;
350	}
351
352	/* Calculate the address of the first meta-block. */
353	if (realbn >= 0)
354		metalbn = -(realbn - bn + NIADDR - i);
355	else
356		metalbn = -(-realbn - bn + NIADDR - i);
357
358	/*
359	 * At each iteration, off is the offset into the bap array which is
360	 * an array of disk addresses at the current level of indirection.
361	 * The logical block number and the offset in that block are stored
362	 * into the argument array.
363	 */
364	ap->in_lbn = metalbn;
365	ap->in_off = off = NIADDR - i;
366	ap++;
367	for (++numlevels; i <= NIADDR; i++) {
368		/* If searching for a meta-data block, quit when found. */
369		if (metalbn == realbn)
370			break;
371
372		off = (bn / blockcnt) % MNINDIR(ump);
373
374		++numlevels;
375		ap->in_lbn = metalbn;
376		ap->in_off = off;
377		++ap;
378
379		metalbn -= -1 + off * blockcnt;
380		blockcnt /= MNINDIR(ump);
381	}
382	if (nump)
383		*nump = numlevels;
384	return (0);
385}
386