xref: /netbsd/sys/lib/libsa/ext2fs.c (revision 6550d01e)
1 /*	$NetBSD: ext2fs.c,v 1.9 2011/01/02 21:37:01 jakllsch Exp $	*/
2 
3 /*
4  * Copyright (c) 1997 Manuel Bouyer.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  */
26 
27 /*-
28  * Copyright (c) 1993
29  *	The Regents of the University of California.  All rights reserved.
30  *
31  * This code is derived from software contributed to Berkeley by
32  * The Mach Operating System project at Carnegie-Mellon University.
33  *
34  * Redistribution and use in source and binary forms, with or without
35  * modification, are permitted provided that the following conditions
36  * are met:
37  * 1. Redistributions of source code must retain the above copyright
38  *    notice, this list of conditions and the following disclaimer.
39  * 2. Redistributions in binary form must reproduce the above copyright
40  *    notice, this list of conditions and the following disclaimer in the
41  *    documentation and/or other materials provided with the distribution.
42  * 3. Neither the name of the University nor the names of its contributors
43  *    may be used to endorse or promote products derived from this software
44  *    without specific prior written permission.
45  *
46  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
47  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
48  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
49  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
50  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
51  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
52  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
53  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
54  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
55  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
56  * SUCH DAMAGE.
57  *
58  *
59  * Copyright (c) 1990, 1991 Carnegie Mellon University
60  * All Rights Reserved.
61  *
62  * Author: David Golub
63  *
64  * Permission to use, copy, modify and distribute this software and its
65  * documentation is hereby granted, provided that both the copyright
66  * notice and this permission notice appear in all copies of the
67  * software, derivative works or modified versions, and any portions
68  * thereof, and that both notices appear in supporting documentation.
69  *
70  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
71  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
72  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
73  *
74  * Carnegie Mellon requests users of this software to return to
75  *
76  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
77  *  School of Computer Science
78  *  Carnegie Mellon University
79  *  Pittsburgh PA 15213-3890
80  *
81  * any improvements or extensions that they make and grant Carnegie the
82  * rights to redistribute these changes.
83  */
84 
85 /*
86  *	Stand-alone file reading package for Ext2 file system.
87  */
88 
89 /* #define EXT2FS_DEBUG */
90 
91 #include <sys/param.h>
92 #include <sys/time.h>
93 #include <ufs/ext2fs/ext2fs_dinode.h>
94 #include <ufs/ext2fs/ext2fs_dir.h>
95 #include <ufs/ext2fs/ext2fs.h>
96 #ifdef _STANDALONE
97 #include <lib/libkern/libkern.h>
98 #else
99 #include <string.h>
100 #endif
101 
102 #include "stand.h"
103 #include "ext2fs.h"
104 
105 #if defined(LIBSA_FS_SINGLECOMPONENT) && !defined(LIBSA_NO_FS_SYMLINK)
106 #define LIBSA_NO_FS_SYMLINK
107 #endif
108 
109 #if defined(LIBSA_NO_TWIDDLE)
110 #define twiddle()
111 #endif
112 
113 #ifndef indp_t
114 #define indp_t		int32_t
115 #endif
116 typedef uint32_t	ino32_t;
117 #ifndef FSBTODB
118 #define FSBTODB(fs, indp) fsbtodb(fs, indp)
119 #endif
120 
121 /*
122  * To avoid having a lot of filesystem-block sized buffers lurking (which
123  * could be 32k) we only keep a few entries of the indirect block map.
124  * With 8k blocks, 2^8 blocks is ~500k so we reread the indirect block
125  * ~13 times pulling in a 6M kernel.
126  * The cache size must be smaller than the smallest filesystem block,
127  * so LN2_IND_CACHE_SZ <= 9 (UFS2 and 4k blocks).
128  */
129 #define LN2_IND_CACHE_SZ	6
130 #define IND_CACHE_SZ		(1 << LN2_IND_CACHE_SZ)
131 #define IND_CACHE_MASK		(IND_CACHE_SZ - 1)
132 
133 /*
134  * In-core open file.
135  */
136 struct file {
137 	off_t		f_seekp;	/* seek pointer */
138 	struct m_ext2fs	*f_fs;		/* pointer to super-block */
139 	struct ext2fs_dinode	f_di;		/* copy of on-disk inode */
140 	uint		f_nishift;	/* for blocks in indirect block */
141 	indp_t		f_ind_cache_block;
142 	indp_t		f_ind_cache[IND_CACHE_SZ];
143 
144 	char		*f_buf;		/* buffer for data block */
145 	size_t		f_buf_size;	/* size of data block */
146 	daddr_t		f_buf_blkno;	/* block number of data block */
147 };
148 
149 static int read_inode(ino32_t, struct open_file *);
150 static int block_map(struct open_file *, indp_t, indp_t *);
151 static int buf_read_file(struct open_file *, char **, size_t *);
152 static int search_directory(const char *, int, struct open_file *, ino32_t *);
153 static int read_sblock(struct open_file *, struct m_ext2fs *);
154 static int read_gdblock(struct open_file *, struct m_ext2fs *);
155 #ifdef EXT2FS_DEBUG
156 static void dump_sblock(struct m_ext2fs *);
157 #endif
158 
159 /*
160  * Read a new inode into a file structure.
161  */
162 static int
163 read_inode(ino32_t inumber, struct open_file *f)
164 {
165 	struct file *fp = (struct file *)f->f_fsdata;
166 	struct m_ext2fs *fs = fp->f_fs;
167 	char *buf;
168 	size_t rsize;
169 	int rc;
170 	daddr_t inode_sector;
171 	struct ext2fs_dinode *dip;
172 
173 	inode_sector = FSBTODB(fs, ino_to_fsba(fs, inumber));
174 
175 	/*
176 	 * Read inode and save it.
177 	 */
178 	buf = fp->f_buf;
179 	twiddle();
180 	rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
181 	    inode_sector, fs->e2fs_bsize, buf, &rsize);
182 	if (rc)
183 		return rc;
184 	if (rsize != fs->e2fs_bsize)
185 		return EIO;
186 
187 	dip = (struct ext2fs_dinode *)(buf +
188 	    EXT2_DINODE_SIZE(fs) * ino_to_fsbo(fs, inumber));
189 	e2fs_iload(dip, &fp->f_di);
190 
191 	/*
192 	 * Clear out the old buffers
193 	 */
194 	fp->f_ind_cache_block = ~0;
195 	fp->f_buf_blkno = -1;
196 	return rc;
197 }
198 
199 /*
200  * Given an offset in a file, find the disk block number that
201  * contains that block.
202  */
203 static int
204 block_map(struct open_file *f, indp_t file_block, indp_t *disk_block_p)
205 {
206 	struct file *fp = (struct file *)f->f_fsdata;
207 	struct m_ext2fs *fs = fp->f_fs;
208 	uint level;
209 	indp_t ind_cache;
210 	indp_t ind_block_num;
211 	size_t rsize;
212 	int rc;
213 	indp_t *buf = (void *)fp->f_buf;
214 
215 	/*
216 	 * Index structure of an inode:
217 	 *
218 	 * e2di_blocks[0..NDADDR-1]
219 	 *			hold block numbers for blocks
220 	 *			0..NDADDR-1
221 	 *
222 	 * e2di_blocks[NDADDR+0]
223 	 *			block NDADDR+0 is the single indirect block
224 	 *			holds block numbers for blocks
225 	 *			NDADDR .. NDADDR + NINDIR(fs)-1
226 	 *
227 	 * e2di_blocks[NDADDR+1]
228 	 *			block NDADDR+1 is the double indirect block
229 	 *			holds block numbers for INDEX blocks for blocks
230 	 *			NDADDR + NINDIR(fs) ..
231 	 *			NDADDR + NINDIR(fs) + NINDIR(fs)**2 - 1
232 	 *
233 	 * e2di_blocks[NDADDR+2]
234 	 *			block NDADDR+2 is the triple indirect block
235 	 *			holds block numbers for	double-indirect
236 	 *			blocks for blocks
237 	 *			NDADDR + NINDIR(fs) + NINDIR(fs)**2 ..
238 	 *			NDADDR + NINDIR(fs) + NINDIR(fs)**2
239 	 *				+ NINDIR(fs)**3 - 1
240 	 */
241 
242 	if (file_block < NDADDR) {
243 		/* Direct block. */
244 		*disk_block_p = fs2h32(fp->f_di.e2di_blocks[file_block]);
245 		return 0;
246 	}
247 
248 	file_block -= NDADDR;
249 
250 	ind_cache = file_block >> LN2_IND_CACHE_SZ;
251 	if (ind_cache == fp->f_ind_cache_block) {
252 		*disk_block_p =
253 		    fs2h32(fp->f_ind_cache[file_block & IND_CACHE_MASK]);
254 		return 0;
255 	}
256 
257 	for (level = 0;;) {
258 		level += fp->f_nishift;
259 		if (file_block < (indp_t)1 << level)
260 			break;
261 		if (level > NIADDR * fp->f_nishift)
262 			/* Block number too high */
263 			return EFBIG;
264 		file_block -= (indp_t)1 << level;
265 	}
266 
267 	ind_block_num =
268 	    fs2h32(fp->f_di.e2di_blocks[NDADDR + (level / fp->f_nishift - 1)]);
269 
270 	for (;;) {
271 		level -= fp->f_nishift;
272 		if (ind_block_num == 0) {
273 			*disk_block_p = 0;	/* missing */
274 			return 0;
275 		}
276 
277 		twiddle();
278 		/*
279 		 * If we were feeling brave, we could work out the number
280 		 * of the disk sector and read a single disk sector instead
281 		 * of a filesystem block.
282 		 * However we don't do this very often anyway...
283 		 */
284 		rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
285 			FSBTODB(fp->f_fs, ind_block_num), fs->e2fs_bsize,
286 			buf, &rsize);
287 		if (rc)
288 			return rc;
289 		if (rsize != fs->e2fs_bsize)
290 			return EIO;
291 		ind_block_num = fs2h32(buf[file_block >> level]);
292 		if (level == 0)
293 			break;
294 		file_block &= (1 << level) - 1;
295 	}
296 
297 	/* Save the part of the block that contains this sector */
298 	memcpy(fp->f_ind_cache, &buf[file_block & ~IND_CACHE_MASK],
299 	    IND_CACHE_SZ * sizeof fp->f_ind_cache[0]);
300 	fp->f_ind_cache_block = ind_cache;
301 
302 	*disk_block_p = ind_block_num;
303 
304 	return 0;
305 }
306 
307 /*
308  * Read a portion of a file into an internal buffer.
309  * Return the location in the buffer and the amount in the buffer.
310  */
311 static int
312 buf_read_file(struct open_file *f, char **buf_p, size_t *size_p)
313 {
314 	struct file *fp = (struct file *)f->f_fsdata;
315 	struct m_ext2fs *fs = fp->f_fs;
316 	long off;
317 	indp_t file_block;
318 	indp_t disk_block;
319 	size_t block_size;
320 	int rc;
321 
322 	off = blkoff(fs, fp->f_seekp);
323 	file_block = lblkno(fs, fp->f_seekp);
324 	block_size = fs->e2fs_bsize;	/* no fragment */
325 
326 	if (file_block != fp->f_buf_blkno) {
327 		rc = block_map(f, file_block, &disk_block);
328 		if (rc)
329 			return rc;
330 
331 		if (disk_block == 0) {
332 			memset(fp->f_buf, 0, block_size);
333 			fp->f_buf_size = block_size;
334 		} else {
335 			twiddle();
336 			rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
337 				FSBTODB(fs, disk_block),
338 				block_size, fp->f_buf, &fp->f_buf_size);
339 			if (rc)
340 				return rc;
341 		}
342 
343 		fp->f_buf_blkno = file_block;
344 	}
345 
346 	/*
347 	 * Return address of byte in buffer corresponding to
348 	 * offset, and size of remainder of buffer after that
349 	 * byte.
350 	 */
351 	*buf_p = fp->f_buf + off;
352 	*size_p = block_size - off;
353 
354 	/*
355 	 * But truncate buffer at end of file.
356 	 */
357 	/* XXX should handle LARGEFILE */
358 	if (*size_p > fp->f_di.e2di_size - fp->f_seekp)
359 		*size_p = fp->f_di.e2di_size - fp->f_seekp;
360 
361 	return 0;
362 }
363 
364 /*
365  * Search a directory for a name and return its
366  * inode number.
367  */
368 static int
369 search_directory(const char *name, int length, struct open_file *f,
370 	ino32_t *inumber_p)
371 {
372 	struct file *fp = (struct file *)f->f_fsdata;
373 	struct ext2fs_direct *dp;
374 	struct ext2fs_direct *edp;
375 	char *buf;
376 	size_t buf_size;
377 	int namlen;
378 	int rc;
379 
380 	fp->f_seekp = 0;
381 	/* XXX should handle LARGEFILE */
382 	while (fp->f_seekp < (off_t)fp->f_di.e2di_size) {
383 		rc = buf_read_file(f, &buf, &buf_size);
384 		if (rc)
385 			return rc;
386 
387 		dp = (struct ext2fs_direct *)buf;
388 		edp = (struct ext2fs_direct *)(buf + buf_size);
389 		for (; dp < edp;
390 		    dp = (void *)((char *)dp + fs2h16(dp->e2d_reclen))) {
391 			if (fs2h16(dp->e2d_reclen) <= 0)
392 				break;
393 			if (fs2h32(dp->e2d_ino) == (ino32_t)0)
394 				continue;
395 			namlen = dp->e2d_namlen;
396 			if (namlen == length &&
397 			    !memcmp(name, dp->e2d_name, length)) {
398 				/* found entry */
399 				*inumber_p = fs2h32(dp->e2d_ino);
400 				return 0;
401 			}
402 		}
403 		fp->f_seekp += buf_size;
404 	}
405 	return ENOENT;
406 }
407 
408 int
409 read_sblock(struct open_file *f, struct m_ext2fs *fs)
410 {
411 	static uint8_t sbbuf[SBSIZE];
412 	struct ext2fs ext2fs;
413 	size_t buf_size;
414 	int rc;
415 
416 	rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
417 	    SBOFF / DEV_BSIZE, SBSIZE, sbbuf, &buf_size);
418 	if (rc)
419 		return rc;
420 
421 	if (buf_size != SBSIZE)
422 		return EIO;
423 
424 	e2fs_sbload((void *)sbbuf, &ext2fs);
425 	if (ext2fs.e2fs_magic != E2FS_MAGIC)
426 		return EINVAL;
427 	if (ext2fs.e2fs_rev > E2FS_REV1 ||
428 	    (ext2fs.e2fs_rev == E2FS_REV1 &&
429 	     (ext2fs.e2fs_first_ino != EXT2_FIRSTINO ||
430 	     (ext2fs.e2fs_inode_size != 128 && ext2fs.e2fs_inode_size != 256) ||
431 	      ext2fs.e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP))) {
432 		return ENODEV;
433 	}
434 
435 	e2fs_sbload((void *)sbbuf, &fs->e2fs);
436 	/* compute in-memory m_ext2fs values */
437 	fs->e2fs_ncg =
438 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
439 	    fs->e2fs.e2fs_bpg);
440 	/* XXX assume hw bsize = 512 */
441 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
442 	fs->e2fs_bsize = MINBSIZE << fs->e2fs.e2fs_log_bsize;
443 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
444 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
445 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
446 	fs->e2fs_ngdb =
447 	    howmany(fs->e2fs_ncg, fs->e2fs_bsize / sizeof(struct ext2_gd));
448 	fs->e2fs_ipb = fs->e2fs_bsize / ext2fs.e2fs_inode_size;
449 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg / fs->e2fs_ipb;
450 
451 	return 0;
452 }
453 
454 int
455 read_gdblock(struct open_file *f, struct m_ext2fs *fs)
456 {
457 	struct file *fp = (struct file *)f->f_fsdata;
458 	size_t rsize;
459 	uint gdpb;
460 	int i, rc;
461 
462 	gdpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
463 
464 	for (i = 0; i < fs->e2fs_ngdb; i++) {
465 		rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
466 		    FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
467 		    1 /* superblock */ + i),
468 		    fs->e2fs_bsize, fp->f_buf, &rsize);
469 		if (rc)
470 			return rc;
471 		if (rsize != fs->e2fs_bsize)
472 			return EIO;
473 
474 		e2fs_cgload((struct ext2_gd *)fp->f_buf,
475 		    &fs->e2fs_gd[i * gdpb],
476 		    (i == (fs->e2fs_ngdb - 1)) ?
477 		    (fs->e2fs_ncg - gdpb * i) * sizeof(struct ext2_gd):
478 		    fs->e2fs_bsize);
479 	}
480 
481 	return 0;
482 }
483 
484 
485 /*
486  * Open a file.
487  */
488 int
489 ext2fs_open(const char *path, struct open_file *f)
490 {
491 #ifndef LIBSA_FS_SINGLECOMPONENT
492 	const char *cp, *ncp;
493 	int c;
494 #endif
495 	ino32_t inumber;
496 	struct file *fp;
497 	struct m_ext2fs *fs;
498 	int rc;
499 #ifndef LIBSA_NO_FS_SYMLINK
500 	ino32_t parent_inumber;
501 	int nlinks = 0;
502 	char namebuf[MAXPATHLEN+1];
503 	char *buf;
504 #endif
505 
506 	/* allocate file system specific data structure */
507 	fp = alloc(sizeof(struct file));
508 	memset(fp, 0, sizeof(struct file));
509 	f->f_fsdata = (void *)fp;
510 
511 	/* allocate space and read super block */
512 	fs = alloc(sizeof(*fs));
513 	memset(fs, 0, sizeof(*fs));
514 	fp->f_fs = fs;
515 	twiddle();
516 
517 	rc = read_sblock(f, fs);
518 	if (rc)
519 		goto out;
520 
521 #ifdef EXT2FS_DEBUG
522 	dump_sblock(fs);
523 #endif
524 
525 	/* alloc a block sized buffer used for all fs transfers */
526 	fp->f_buf = alloc(fs->e2fs_bsize);
527 
528 	/* read group descriptor blocks */
529 	fs->e2fs_gd = alloc(sizeof(struct ext2_gd) * fs->e2fs_ncg);
530 	rc = read_gdblock(f, fs);
531 	if (rc)
532 		goto out;
533 
534 	/*
535 	 * Calculate indirect block levels.
536 	 */
537 	{
538 		indp_t mult;
539 		int ln2;
540 
541 		/*
542 		 * We note that the number of indirect blocks is always
543 		 * a power of 2.  This lets us use shifts and masks instead
544 		 * of divide and remainder and avoinds pulling in the
545 		 * 64bit division routine into the boot code.
546 		 */
547 		mult = NINDIR(fs);
548 #ifdef DEBUG
549 		if (!powerof2(mult)) {
550 			/* Hummm was't a power of 2 */
551 			rc = EINVAL;
552 			goto out;
553 		}
554 #endif
555 		for (ln2 = 0; mult != 1; ln2++)
556 			mult >>= 1;
557 
558 		fp->f_nishift = ln2;
559 	}
560 
561 	inumber = EXT2_ROOTINO;
562 	if ((rc = read_inode(inumber, f)) != 0)
563 		goto out;
564 
565 #ifndef LIBSA_FS_SINGLECOMPONENT
566 	cp = path;
567 	while (*cp) {
568 
569 		/*
570 		 * Remove extra separators
571 		 */
572 		while (*cp == '/')
573 			cp++;
574 		if (*cp == '\0')
575 			break;
576 
577 		/*
578 		 * Check that current node is a directory.
579 		 */
580 		if ((fp->f_di.e2di_mode & EXT2_IFMT) != EXT2_IFDIR) {
581 			rc = ENOTDIR;
582 			goto out;
583 		}
584 
585 		/*
586 		 * Get next component of path name.
587 		 */
588 		ncp = cp;
589 		while ((c = *cp) != '\0' && c != '/')
590 			cp++;
591 
592 		/*
593 		 * Look up component in current directory.
594 		 * Save directory inumber in case we find a
595 		 * symbolic link.
596 		 */
597 #ifndef LIBSA_NO_FS_SYMLINK
598 		parent_inumber = inumber;
599 #endif
600 		rc = search_directory(ncp, cp - ncp, f, &inumber);
601 		if (rc)
602 			goto out;
603 
604 		/*
605 		 * Open next component.
606 		 */
607 		if ((rc = read_inode(inumber, f)) != 0)
608 			goto out;
609 
610 #ifndef LIBSA_NO_FS_SYMLINK
611 		/*
612 		 * Check for symbolic link.
613 		 */
614 		if ((fp->f_di.e2di_mode & EXT2_IFMT) == EXT2_IFLNK) {
615 			/* XXX should handle LARGEFILE */
616 			int link_len = fp->f_di.e2di_size;
617 			int len;
618 
619 			len = strlen(cp);
620 
621 			if (link_len + len > MAXPATHLEN ||
622 			    ++nlinks > MAXSYMLINKS) {
623 				rc = ENOENT;
624 				goto out;
625 			}
626 
627 			memmove(&namebuf[link_len], cp, len + 1);
628 
629 			if (link_len < EXT2_MAXSYMLINKLEN) {
630 				memcpy(namebuf, fp->f_di.e2di_blocks, link_len);
631 			} else {
632 				/*
633 				 * Read file for symbolic link
634 				 */
635 				size_t buf_size;
636 				indp_t	disk_block;
637 
638 				buf = fp->f_buf;
639 				rc = block_map(f, (indp_t)0, &disk_block);
640 				if (rc)
641 					goto out;
642 
643 				twiddle();
644 				rc = DEV_STRATEGY(f->f_dev)(f->f_devdata,
645 					F_READ, FSBTODB(fs, disk_block),
646 					fs->e2fs_bsize, buf, &buf_size);
647 				if (rc)
648 					goto out;
649 
650 				memcpy(namebuf, buf, link_len);
651 			}
652 
653 			/*
654 			 * If relative pathname, restart at parent directory.
655 			 * If absolute pathname, restart at root.
656 			 */
657 			cp = namebuf;
658 			if (*cp != '/')
659 				inumber = parent_inumber;
660 			else
661 				inumber = (ino32_t)EXT2_ROOTINO;
662 
663 			if ((rc = read_inode(inumber, f)) != 0)
664 				goto out;
665 		}
666 #endif	/* !LIBSA_NO_FS_SYMLINK */
667 	}
668 
669 	/*
670 	 * Found terminal component.
671 	 */
672 	rc = 0;
673 
674 #else /* !LIBSA_FS_SINGLECOMPONENT */
675 
676 	/* look up component in the current (root) directory */
677 	rc = search_directory(path, strlen(path), f, &inumber);
678 	if (rc)
679 		goto out;
680 
681 	/* open it */
682 	rc = read_inode(inumber, f);
683 
684 #endif /* !LIBSA_FS_SINGLECOMPONENT */
685 
686 	fp->f_seekp = 0;		/* reset seek pointer */
687 
688 out:
689 	if (rc)
690 		ext2fs_close(f);
691 	else {
692 		fsmod = "ext2fs";
693 		fsmod2 = "ffs";
694 	}
695 	return rc;
696 }
697 
698 int
699 ext2fs_close(struct open_file *f)
700 {
701 	struct file *fp = (struct file *)f->f_fsdata;
702 
703 	f->f_fsdata = NULL;
704 	if (fp == NULL)
705 		return 0;
706 
707 	if (fp->f_fs->e2fs_gd)
708 		dealloc(fp->f_fs->e2fs_gd,
709 		    sizeof(struct ext2_gd) * fp->f_fs->e2fs_ncg);
710 	if (fp->f_buf)
711 		dealloc(fp->f_buf, fp->f_fs->e2fs_bsize);
712 	dealloc(fp->f_fs, sizeof(*fp->f_fs));
713 	dealloc(fp, sizeof(struct file));
714 	return 0;
715 }
716 
717 /*
718  * Copy a portion of a file into kernel memory.
719  * Cross block boundaries when necessary.
720  */
721 int
722 ext2fs_read(struct open_file *f, void *start, size_t size, size_t *resid)
723 {
724 	struct file *fp = (struct file *)f->f_fsdata;
725 	size_t csize;
726 	char *buf;
727 	size_t buf_size;
728 	int rc = 0;
729 	char *addr = start;
730 
731 	while (size != 0) {
732 		/* XXX should handle LARGEFILE */
733 		if (fp->f_seekp >= (off_t)fp->f_di.e2di_size)
734 			break;
735 
736 		rc = buf_read_file(f, &buf, &buf_size);
737 		if (rc)
738 			break;
739 
740 		csize = size;
741 		if (csize > buf_size)
742 			csize = buf_size;
743 
744 		memcpy(addr, buf, csize);
745 
746 		fp->f_seekp += csize;
747 		addr += csize;
748 		size -= csize;
749 	}
750 	if (resid)
751 		*resid = size;
752 	return rc;
753 }
754 
755 /*
756  * Not implemented.
757  */
758 #ifndef LIBSA_NO_FS_WRITE
759 int
760 ext2fs_write(struct open_file *f, void *start, size_t size, size_t *resid)
761 {
762 
763 	return EROFS;
764 }
765 #endif /* !LIBSA_NO_FS_WRITE */
766 
767 #ifndef LIBSA_NO_FS_SEEK
768 off_t
769 ext2fs_seek(struct open_file *f, off_t offset, int where)
770 {
771 	struct file *fp = (struct file *)f->f_fsdata;
772 
773 	switch (where) {
774 	case SEEK_SET:
775 		fp->f_seekp = offset;
776 		break;
777 	case SEEK_CUR:
778 		fp->f_seekp += offset;
779 		break;
780 	case SEEK_END:
781 		/* XXX should handle LARGEFILE */
782 		fp->f_seekp = fp->f_di.e2di_size - offset;
783 		break;
784 	default:
785 		return -1;
786 	}
787 	return fp->f_seekp;
788 }
789 #endif /* !LIBSA_NO_FS_SEEK */
790 
791 int
792 ext2fs_stat(struct open_file *f, struct stat *sb)
793 {
794 	struct file *fp = (struct file *)f->f_fsdata;
795 
796 	/* only important stuff */
797 	memset(sb, 0, sizeof *sb);
798 	sb->st_mode = fp->f_di.e2di_mode;
799 	sb->st_uid = fp->f_di.e2di_uid;
800 	sb->st_gid = fp->f_di.e2di_gid;
801 	/* XXX should handle LARGEFILE */
802 	sb->st_size = fp->f_di.e2di_size;
803 	return 0;
804 }
805 
806 /*
807  * byte swap functions for big endian machines
808  * (ext2fs is always little endian)
809  *
810  * XXX: We should use src/sys/ufs/ext2fs/ext2fs_bswap.c
811  */
812 
813 /* These functions are only needed if native byte order is not big endian */
814 #if BYTE_ORDER == BIG_ENDIAN
815 void
816 e2fs_sb_bswap(struct ext2fs *old, struct ext2fs *new)
817 {
818 
819 	/* preserve unused fields */
820 	memcpy(new, old, sizeof(struct ext2fs));
821 	new->e2fs_icount	=	bswap32(old->e2fs_icount);
822 	new->e2fs_bcount	=	bswap32(old->e2fs_bcount);
823 	new->e2fs_rbcount	=	bswap32(old->e2fs_rbcount);
824 	new->e2fs_fbcount	=	bswap32(old->e2fs_fbcount);
825 	new->e2fs_ficount	=	bswap32(old->e2fs_ficount);
826 	new->e2fs_first_dblock	=	bswap32(old->e2fs_first_dblock);
827 	new->e2fs_log_bsize	=	bswap32(old->e2fs_log_bsize);
828 	new->e2fs_fsize		=	bswap32(old->e2fs_fsize);
829 	new->e2fs_bpg		=	bswap32(old->e2fs_bpg);
830 	new->e2fs_fpg		=	bswap32(old->e2fs_fpg);
831 	new->e2fs_ipg		=	bswap32(old->e2fs_ipg);
832 	new->e2fs_mtime		=	bswap32(old->e2fs_mtime);
833 	new->e2fs_wtime		=	bswap32(old->e2fs_wtime);
834 	new->e2fs_mnt_count	=	bswap16(old->e2fs_mnt_count);
835 	new->e2fs_max_mnt_count	=	bswap16(old->e2fs_max_mnt_count);
836 	new->e2fs_magic		=	bswap16(old->e2fs_magic);
837 	new->e2fs_state		=	bswap16(old->e2fs_state);
838 	new->e2fs_beh		=	bswap16(old->e2fs_beh);
839 	new->e2fs_minrev	=	bswap16(old->e2fs_minrev);
840 	new->e2fs_lastfsck	=	bswap32(old->e2fs_lastfsck);
841 	new->e2fs_fsckintv	=	bswap32(old->e2fs_fsckintv);
842 	new->e2fs_creator	=	bswap32(old->e2fs_creator);
843 	new->e2fs_rev		=	bswap32(old->e2fs_rev);
844 	new->e2fs_ruid		=	bswap16(old->e2fs_ruid);
845 	new->e2fs_rgid		=	bswap16(old->e2fs_rgid);
846 	new->e2fs_first_ino	=	bswap32(old->e2fs_first_ino);
847 	new->e2fs_inode_size	=	bswap16(old->e2fs_inode_size);
848 	new->e2fs_block_group_nr =	bswap16(old->e2fs_block_group_nr);
849 	new->e2fs_features_compat =	bswap32(old->e2fs_features_compat);
850 	new->e2fs_features_incompat =	bswap32(old->e2fs_features_incompat);
851 	new->e2fs_features_rocompat =	bswap32(old->e2fs_features_rocompat);
852 	new->e2fs_algo		=	bswap32(old->e2fs_algo);
853 	new->e2fs_reserved_ngdb	=	bswap16(old->e2fs_reserved_ngdb);
854 }
855 
856 void e2fs_cg_bswap(struct ext2_gd *old, struct ext2_gd *new, int size)
857 {
858 	int i;
859 
860 	for (i = 0; i < (size / sizeof(struct ext2_gd)); i++) {
861 		new[i].ext2bgd_b_bitmap	= bswap32(old[i].ext2bgd_b_bitmap);
862 		new[i].ext2bgd_i_bitmap	= bswap32(old[i].ext2bgd_i_bitmap);
863 		new[i].ext2bgd_i_tables	= bswap32(old[i].ext2bgd_i_tables);
864 		new[i].ext2bgd_nbfree	= bswap16(old[i].ext2bgd_nbfree);
865 		new[i].ext2bgd_nifree	= bswap16(old[i].ext2bgd_nifree);
866 		new[i].ext2bgd_ndirs	= bswap16(old[i].ext2bgd_ndirs);
867 	}
868 }
869 
870 void e2fs_i_bswap(struct ext2fs_dinode *old, struct ext2fs_dinode *new)
871 {
872 
873 	new->e2di_mode		=	bswap16(old->e2di_mode);
874 	new->e2di_uid		=	bswap16(old->e2di_uid);
875 	new->e2di_gid		=	bswap16(old->e2di_gid);
876 	new->e2di_nlink		=	bswap16(old->e2di_nlink);
877 	new->e2di_size		=	bswap32(old->e2di_size);
878 	new->e2di_atime		=	bswap32(old->e2di_atime);
879 	new->e2di_ctime		=	bswap32(old->e2di_ctime);
880 	new->e2di_mtime		=	bswap32(old->e2di_mtime);
881 	new->e2di_dtime		=	bswap32(old->e2di_dtime);
882 	new->e2di_nblock	=	bswap32(old->e2di_nblock);
883 	new->e2di_flags		=	bswap32(old->e2di_flags);
884 	new->e2di_gen		=	bswap32(old->e2di_gen);
885 	new->e2di_facl		=	bswap32(old->e2di_facl);
886 	new->e2di_dacl		=	bswap32(old->e2di_dacl);
887 	new->e2di_faddr		=	bswap32(old->e2di_faddr);
888 	memcpy(&new->e2di_blocks[0], &old->e2di_blocks[0],
889 	    (NDADDR + NIADDR) * sizeof(uint32_t));
890 }
891 #endif
892 
893 #ifdef EXT2FS_DEBUG
894 void
895 dump_sblock(struct m_ext2fs *fs)
896 {
897 
898 	printf("fs->e2fs.e2fs_bcount = %u\n", fs->e2fs.e2fs_bcount);
899 	printf("fs->e2fs.e2fs_first_dblock = %u\n", fs->e2fs.e2fs_first_dblock);
900 	printf("fs->e2fs.e2fs_log_bsize = %u\n", fs->e2fs.e2fs_log_bsize);
901 	printf("fs->e2fs.e2fs_bpg = %u\n", fs->e2fs.e2fs_bpg);
902 	printf("fs->e2fs.e2fs_ipg = %u\n", fs->e2fs.e2fs_ipg);
903 	printf("fs->e2fs.e2fs_magic = 0x%x\n", fs->e2fs.e2fs_magic);
904 	printf("fs->e2fs.e2fs_rev = %u\n", fs->e2fs.e2fs_rev);
905 
906 	if (fs->e2fs.e2fs_rev == E2FS_REV1) {
907 		printf("fs->e2fs.e2fs_first_ino = %u\n",
908 		    fs->e2fs.e2fs_first_ino);
909 		printf("fs->e2fs.e2fs_inode_size = %u\n",
910 		    fs->e2fs.e2fs_inode_size);
911 		printf("fs->e2fs.e2fs_features_compat = %u\n",
912 		    fs->e2fs.e2fs_features_compat);
913 		printf("fs->e2fs.e2fs_features_incompat = %u\n",
914 		    fs->e2fs.e2fs_features_incompat);
915 		printf("fs->e2fs.e2fs_features_rocompat = %u\n",
916 		    fs->e2fs.e2fs_features_rocompat);
917 		printf("fs->e2fs.e2fs_reserved_ngdb = %u\n",
918 		    fs->e2fs.e2fs_reserved_ngdb);
919 	}
920 
921 	printf("fs->e2fs_bsize = %u\n", fs->e2fs_bsize);
922 	printf("fs->e2fs_fsbtodb = %u\n", fs->e2fs_fsbtodb);
923 	printf("fs->e2fs_ncg = %u\n", fs->e2fs_ncg);
924 	printf("fs->e2fs_ngdb = %u\n", fs->e2fs_ngdb);
925 	printf("fs->e2fs_ipb = %u\n", fs->e2fs_ipb);
926 	printf("fs->e2fs_itpg = %u\n", fs->e2fs_itpg);
927 }
928 #endif
929