xref: /freebsd/stand/libsa/ext2fs.c (revision 2e3f49888ec8851bafb22011533217487764fdb0)
1 /*-
2  * Copyright (c) 1999,2000 Jonathan Lemon <jlemon@freebsd.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 /*-
29  * Copyright (c) 1993
30  *	The Regents of the University of California.  All rights reserved.
31  *
32  * This code is derived from software contributed to Berkeley by
33  * The Mach Operating System project at Carnegie-Mellon University.
34  *
35  * Redistribution and use in source and binary forms, with or without
36  * modification, are permitted provided that the following conditions
37  * are met:
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgement:
45  *	This product includes software developed by the University of
46  *	California, Berkeley and its contributors.
47  * 4. Neither the name of the University nor the names of its contributors
48  *    may be used to endorse or promote products derived from this software
49  *    without specific prior written permission.
50  *
51  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61  * SUCH DAMAGE.
62  *
63  *
64  * Copyright (c) 1990, 1991 Carnegie Mellon University
65  * All Rights Reserved.
66  *
67  * Author: David Golub
68  *
69  * Permission to use, copy, modify and distribute this software and its
70  * documentation is hereby granted, provided that both the copyright
71  * notice and this permission notice appear in all copies of the
72  * software, derivative works or modified versions, and any portions
73  * thereof, and that both notices appear in supporting documentation.
74  *
75  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
76  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
77  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
78  *
79  * Carnegie Mellon requests users of this software to return to
80  *
81  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
82  *  School of Computer Science
83  *  Carnegie Mellon University
84  *  Pittsburgh PA 15213-3890
85  *
86  * any improvements or extensions that they make and grant Carnegie the
87  * rights to redistribute these changes.
88  */
89 
90 #include <sys/param.h>
91 #include <sys/time.h>
92 #include "stand.h"
93 #include "string.h"
94 
95 static int	ext2fs_open(const char *path, struct open_file *f);
96 static int	ext2fs_close(struct open_file *f);
97 static int	ext2fs_read(struct open_file *f, void *buf,
98 			 size_t size, size_t *resid);
99 static off_t	ext2fs_seek(struct open_file *f, off_t offset, int where);
100 static int	ext2fs_stat(struct open_file *f, struct stat *sb);
101 static int	ext2fs_readdir(struct open_file *f, struct dirent *d);
102 
103 static int dtmap[] = { DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR,
104 			 DT_BLK, DT_FIFO, DT_SOCK, DT_LNK };
105 #define EXTFTODT(x)	(x) > sizeof(dtmap) / sizeof(dtmap[0]) ? \
106 			DT_UNKNOWN : dtmap[x]
107 
108 struct fs_ops ext2fs_fsops = {
109 	.fs_name = "ext2fs",
110 	.fo_open = ext2fs_open,
111 	.fo_close = ext2fs_close,
112 	.fo_read = ext2fs_read,
113 	.fo_write = null_write,
114 	.fo_seek = ext2fs_seek,
115 	.fo_stat = ext2fs_stat,
116 	.fo_readdir = ext2fs_readdir,
117 };
118 
119 #define	EXT2_SBSIZE	1024
120 #define	EXT2_SBLOCK	(1024 / DEV_BSIZE)	/* block offset of superblock */
121 #define EXT2_MAGIC	0xef53
122 #define EXT2_ROOTINO	2
123 
124 #define EXT2_REV0		0	/* original revision of ext2 */
125 #define EXT2_R0_ISIZE		128	/* inode size */
126 #define EXT2_R0_FIRSTINO	11	/* first inode */
127 
128 #define EXT2_MINBSHIFT		10	/* minimum block shift */
129 #define EXT2_MINFSHIFT		10	/* minimum frag shift */
130 
131 #define EXT2_NDADDR		12	/* # of direct blocks */
132 #define EXT2_NIADDR		3	/* # of indirect blocks */
133 
134 /*
135  * file system block to disk address
136  */
137 #define fsb_to_db(fs, blk)	((blk) << (fs)->fs_fsbtodb)
138 
139 /*
140  * inode to block group offset
141  * inode to block group
142  * inode to disk address
143  * inode to block offset
144  */
145 #define ino_to_bgo(fs, ino)	(((ino) - 1) % (fs)->fs_ipg)
146 #define ino_to_bg(fs, ino)	(((ino) - 1) / (fs)->fs_ipg)
147 #define ino_to_db(fs, bg, ino) \
148 	fsb_to_db(fs, ((bg)[ino_to_bg(fs, ino)].bg_inotbl + \
149 	    ino_to_bgo(fs, ino) / (fs)->fs_ipb))
150 #define ino_to_bo(fs, ino)	(ino_to_bgo(fs, ino) % (fs)->fs_ipb)
151 
152 #define nindir(fs) \
153 	((fs)->fs_bsize / sizeof(uint32_t))
154 #define lblkno(fs, loc)				/* loc / bsize */ \
155 	((loc) >> (fs)->fs_bshift)
156 #define smalllblktosize(fs, blk)		/* blk * bsize */ \
157 	((blk) << (fs)->fs_bshift)
158 #define blkoff(fs, loc)				/* loc % bsize */ \
159 	((loc) & (fs)->fs_bmask)
160 #define fragroundup(fs, size)			/* roundup(size, fsize) */ \
161 	(((size) + (fs)->fs_fmask) & ~(fs)->fs_fmask)
162 #define dblksize(fs, dip, lbn) \
163 	(((lbn) >= EXT2_NDADDR || (dip)->di_size >= smalllblktosize(fs, (lbn) + 1)) \
164 	    ? (fs)->fs_bsize \
165 	    : (fragroundup(fs, blkoff(fs, (dip)->di_size))))
166 
167 /*
168  * superblock describing ext2fs
169  */
170 struct ext2fs_disk {
171 	uint32_t	fd_inodes;	/* # of inodes */
172 	uint32_t	fd_blocks;	/* # of blocks */
173 	uint32_t	fd_resblk;	/* # of reserved blocks */
174 	uint32_t	fd_freeblk;	/* # of free blocks */
175 	uint32_t	fd_freeino;	/* # of free inodes */
176 	uint32_t	fd_firstblk;	/* first data block */
177 	uint32_t	fd_bsize;	/* block size */
178 	uint32_t	fd_fsize;	/* frag size */
179 	uint32_t	fd_bpg;		/* blocks per group */
180 	uint32_t	fd_fpg;		/* frags per group */
181 	uint32_t	fd_ipg;		/* inodes per group */
182 	uint32_t	fd_mtime;	/* mount time */
183 	uint32_t	fd_wtime;	/* write time */
184 	uint16_t	fd_mount;	/* # of mounts */
185 	int16_t		fd_maxmount;	/* max # of mounts */
186 	uint16_t	fd_magic;	/* magic number */
187 	uint16_t	fd_state;	/* state */
188 	uint16_t	fd_eflag;	/* error flags */
189 	uint16_t	fd_mnrrev;	/* minor revision */
190 	uint32_t	fd_lastchk;	/* last check */
191 	uint32_t	fd_chkintvl;	/* maximum check interval */
192 	uint32_t	fd_os;		/* os */
193 	uint32_t	fd_revision;	/* revision */
194 	uint16_t	fd_uid;		/* uid for reserved blocks */
195 	uint16_t	fd_gid;		/* gid for reserved blocks */
196 
197 	uint32_t	fd_firstino;	/* first non-reserved inode */
198 	uint16_t	fd_isize;	/* inode size */
199 	uint16_t	fd_nblkgrp;	/* block group # of superblock */
200 	uint32_t	fd_fcompat;	/* compatible features */
201 	uint32_t	fd_fincompat;	/* incompatible features */
202 	uint32_t	fd_frocompat;	/* read-only compatibilties */
203 	uint8_t		fd_uuid[16];	/* volume uuid */
204 	char 		fd_volname[16];	/* volume name */
205 	char 		fd_fsmnt[64];	/* name last mounted on */
206 	uint32_t	fd_bitmap;	/* compression bitmap */
207 
208 	uint8_t		fd_nblkpa;	/* # of blocks to preallocate */
209 	uint8_t		fd_ndblkpa;	/* # of dir blocks to preallocate */
210 };
211 
212 struct ext2fs_core {
213 	int		fc_bsize;	/* block size */
214 	int		fc_bshift;	/* block shift amount */
215 	int		fc_bmask;	/* block mask */
216 	int		fc_fsize;	/* frag size */
217 	int		fc_fshift;	/* frag shift amount */
218 	int		fc_fmask;	/* frag mask */
219 	int		fc_isize;	/* inode size */
220 	int		fc_imask;	/* inode mask */
221 	int		fc_firstino;	/* first non-reserved inode */
222 	int		fc_ipb;		/* inodes per block */
223 	int		fc_fsbtodb;	/* fsb to ds shift */
224 };
225 
226 struct ext2fs {
227 	struct		ext2fs_disk fs_fd;
228 	char		fs_pad[EXT2_SBSIZE - sizeof(struct ext2fs_disk)];
229 	struct		ext2fs_core fs_fc;
230 
231 #define fs_magic	fs_fd.fd_magic
232 #define fs_revision	fs_fd.fd_revision
233 #define fs_blocks	fs_fd.fd_blocks
234 #define fs_firstblk	fs_fd.fd_firstblk
235 #define fs_bpg		fs_fd.fd_bpg
236 #define fs_ipg		fs_fd.fd_ipg
237 
238 #define fs_bsize	fs_fc.fc_bsize
239 #define fs_bshift	fs_fc.fc_bshift
240 #define fs_bmask	fs_fc.fc_bmask
241 #define fs_fsize	fs_fc.fc_fsize
242 #define fs_fshift	fs_fc.fc_fshift
243 #define fs_fmask	fs_fc.fc_fmask
244 #define fs_isize	fs_fc.fc_isize
245 #define fs_imask	fs_fc.fc_imask
246 #define fs_firstino	fs_fc.fc_firstino
247 #define fs_ipb		fs_fc.fc_ipb
248 #define fs_fsbtodb	fs_fc.fc_fsbtodb
249 };
250 
251 struct ext2blkgrp {
252 	uint32_t	bg_blkmap;	/* block bitmap */
253 	uint32_t	bg_inomap;	/* inode bitmap */
254 	uint32_t	bg_inotbl;	/* inode table */
255 	uint16_t	bg_nfblk;	/* # of free blocks */
256 	uint16_t	bg_nfino;	/* # of free inodes */
257 	uint16_t	bg_ndirs;	/* # of dirs */
258 	char		bg_pad[14];
259 };
260 
261 struct ext2dinode {
262 	uint16_t	di_mode;	/* mode */
263 	uint16_t	di_uid;		/* uid */
264 	uint32_t	di_size;	/* byte size */
265 	uint32_t	di_atime;	/* access time */
266 	uint32_t	di_ctime;	/* creation time */
267 	uint32_t	di_mtime;	/* modification time */
268 	uint32_t	di_dtime;	/* deletion time */
269 	uint16_t	di_gid;		/* gid */
270 	uint16_t	di_nlink;	/* link count */
271 	uint32_t	di_nblk;	/* block count */
272 	uint32_t	di_flags;	/* file flags */
273 
274 	uint32_t	di_osdep1;	/* os dependent stuff */
275 
276 	uint32_t	di_db[EXT2_NDADDR]; /* direct blocks */
277 	uint32_t	di_ib[EXT2_NIADDR]; /* indirect blocks */
278 	uint32_t	di_version;	/* version */
279 	uint32_t	di_facl;	/* file acl */
280 	uint32_t	di_dacl;	/* dir acl */
281 	uint32_t	di_faddr;	/* fragment addr */
282 
283 	uint8_t		di_frag;	/* fragment number */
284 	uint8_t		di_fsize;	/* fragment size */
285 
286 	char		di_pad[10];
287 
288 #define di_shortlink	di_db
289 };
290 
291 #define EXT2_MAXNAMLEN       255
292 
293 struct ext2dirent {
294 	uint32_t	d_ino;		/* inode */
295 	uint16_t	d_reclen;	/* directory entry length */
296 	uint8_t		d_namlen;	/* name length */
297 	uint8_t		d_type;		/* file type */
298 	char		d_name[EXT2_MAXNAMLEN];
299 };
300 
301 struct file {
302 	off_t		f_seekp;		/* seek pointer */
303 	struct 		ext2fs *f_fs;		/* pointer to super-block */
304 	struct 		ext2blkgrp *f_bg;	/* pointer to blkgrp map */
305 	struct 		ext2dinode f_di;	/* copy of on-disk inode */
306 	int		f_nindir[EXT2_NIADDR];	/* number of blocks mapped by
307 						   indirect block at level i */
308 	char		*f_blk[EXT2_NIADDR];	/* buffer for indirect block
309 						   at level i */
310 	size_t		f_blksize[EXT2_NIADDR];	/* size of buffer */
311 	daddr_t		f_blkno[EXT2_NIADDR];	/* disk address of block in
312 						   buffer */
313 	char		*f_buf;			/* buffer for data block */
314 	size_t		f_buf_size;		/* size of data block */
315 	daddr_t		f_buf_blkno;		/* block number of data block */
316 };
317 
318 /* forward decls */
319 static int 	read_inode(ino_t inumber, struct open_file *f);
320 static int	block_map(struct open_file *f, daddr_t file_block,
321 		    daddr_t *disk_block_p);
322 static int	buf_read_file(struct open_file *f, char **buf_p,
323 		    size_t *size_p);
324 static int	search_directory(char *name, struct open_file *f,
325 		    ino_t *inumber_p);
326 
327 /*
328  * Open a file.
329  */
330 static int
331 ext2fs_open(const char *upath, struct open_file *f)
332 {
333 	struct file *fp;
334 	struct ext2fs *fs;
335 	size_t buf_size;
336 	ino_t inumber, parent_inumber;
337 	int i, len, groups, bg_per_blk, blkgrps, mult;
338 	int nlinks = 0;
339 	int error = 0;
340 	char *cp, *ncp, *path = NULL, *buf = NULL;
341 	char namebuf[MAXPATHLEN+1];
342 	char c;
343 
344 	/* allocate file system specific data structure */
345 	fp = malloc(sizeof(struct file));
346 	if (fp == NULL)
347 		return (ENOMEM);
348 	bzero(fp, sizeof(struct file));
349 	f->f_fsdata = (void *)fp;
350 
351 	/* allocate space and read super block */
352 	fs = (struct ext2fs *)malloc(sizeof(*fs));
353 	fp->f_fs = fs;
354 	twiddle(1);
355 	error = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
356 	    EXT2_SBLOCK, EXT2_SBSIZE, (char *)fs, &buf_size);
357 	if (error)
358 		goto out;
359 
360 	if (buf_size != EXT2_SBSIZE || fs->fs_magic != EXT2_MAGIC) {
361 		error = EINVAL;
362 		goto out;
363 	}
364 
365 	/*
366 	 * compute in-core values for the superblock
367 	 */
368 	fs->fs_bshift = EXT2_MINBSHIFT + fs->fs_fd.fd_bsize;
369 	fs->fs_bsize = 1 << fs->fs_bshift;
370 	fs->fs_bmask = fs->fs_bsize - 1;
371 
372 	fs->fs_fshift = EXT2_MINFSHIFT + fs->fs_fd.fd_fsize;
373 	fs->fs_fsize = 1 << fs->fs_fshift;
374 	fs->fs_fmask = fs->fs_fsize - 1;
375 
376 	if (fs->fs_revision == EXT2_REV0) {
377 		fs->fs_isize = EXT2_R0_ISIZE;
378 		fs->fs_firstino = EXT2_R0_FIRSTINO;
379 	} else {
380 		fs->fs_isize = fs->fs_fd.fd_isize;
381 		fs->fs_firstino = fs->fs_fd.fd_firstino;
382 	}
383 	fs->fs_imask = fs->fs_isize - 1;
384 	fs->fs_ipb = fs->fs_bsize / fs->fs_isize;
385 	fs->fs_fsbtodb = (fs->fs_bsize / DEV_BSIZE) - 1;
386 
387 	/*
388 	 * we have to load in the "group descriptors" here
389 	 */
390 	groups = howmany(fs->fs_blocks - fs->fs_firstblk, fs->fs_bpg);
391 	bg_per_blk = fs->fs_bsize / sizeof(struct ext2blkgrp);
392 	blkgrps = howmany(groups, bg_per_blk);
393 	len = blkgrps * fs->fs_bsize;
394 
395 	fp->f_bg = malloc(len);
396 	twiddle(1);
397 	error = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
398 	    EXT2_SBLOCK + EXT2_SBSIZE / DEV_BSIZE, len,
399 	    (char *)fp->f_bg, &buf_size);
400 	if (error)
401 		goto out;
402 
403 	/*
404 	 * XXX
405 	 * validation of values?  (blocksize, descriptors, etc?)
406 	 */
407 
408 	/*
409 	 * Calculate indirect block levels.
410 	 */
411 	mult = 1;
412 	for (i = 0; i < EXT2_NIADDR; i++) {
413 		mult *= nindir(fs);
414 		fp->f_nindir[i] = mult;
415 	}
416 
417 	inumber = EXT2_ROOTINO;
418 	if ((error = read_inode(inumber, f)) != 0)
419 		goto out;
420 
421 	path = strdup(upath);
422 	if (path == NULL) {
423 		error = ENOMEM;
424 		goto out;
425 	}
426 	cp = path;
427 	while (*cp) {
428 		/*
429 		 * Remove extra separators
430 		 */
431 		while (*cp == '/')
432 			cp++;
433 		if (*cp == '\0')
434 			break;
435 
436 		/*
437 		 * Check that current node is a directory.
438 		 */
439 		if (! S_ISDIR(fp->f_di.di_mode)) {
440 			error = ENOTDIR;
441 			goto out;
442 		}
443 
444 		/*
445 		 * Get next component of path name.
446 		 */
447 		len = 0;
448 
449 		ncp = cp;
450 		while ((c = *cp) != '\0' && c != '/') {
451 			if (++len > EXT2_MAXNAMLEN) {
452 				error = ENOENT;
453 				goto out;
454 			}
455 			cp++;
456 		}
457 		*cp = '\0';
458 
459 		/*
460 		 * Look up component in current directory.
461 		 * Save directory inumber in case we find a
462 		 * symbolic link.
463 		 */
464 		parent_inumber = inumber;
465 		error = search_directory(ncp, f, &inumber);
466 		*cp = c;
467 		if (error)
468 			goto out;
469 
470 		/*
471 		 * Open next component.
472 		 */
473 		if ((error = read_inode(inumber, f)) != 0)
474 			goto out;
475 
476 		/*
477 		 * Check for symbolic link.
478 		 */
479 		if (S_ISLNK(fp->f_di.di_mode)) {
480 			int link_len = fp->f_di.di_size;
481 			int len;
482 
483 			len = strlen(cp);
484 			if (link_len + len > MAXPATHLEN ||
485 			    ++nlinks > MAXSYMLINKS) {
486 				error = ENOENT;
487 				goto out;
488 			}
489 
490 			bcopy(cp, &namebuf[link_len], len + 1);
491 			if (fp->f_di.di_nblk == 0) {
492 				bcopy(fp->f_di.di_shortlink,
493 				    namebuf, link_len);
494 			} else {
495 				/*
496 				 * Read file for symbolic link
497 				 */
498 				struct ext2fs *fs = fp->f_fs;
499 				daddr_t	disk_block;
500 				size_t buf_size;
501 
502 				if (! buf)
503 					buf = malloc(fs->fs_bsize);
504 				error = block_map(f, (daddr_t)0, &disk_block);
505 				if (error)
506 					goto out;
507 
508 				twiddle(1);
509 				error = (f->f_dev->dv_strategy)(f->f_devdata,
510 				    F_READ, fsb_to_db(fs, disk_block),
511 				    fs->fs_bsize, buf, &buf_size);
512 				if (error)
513 					goto out;
514 
515 				bcopy((char *)buf, namebuf, link_len);
516 			}
517 
518 			/*
519 			 * If relative pathname, restart at parent directory.
520 			 * If absolute pathname, restart at root.
521 			 */
522 			cp = namebuf;
523 			if (*cp != '/')
524 				inumber = parent_inumber;
525 			else
526 				inumber = (ino_t)EXT2_ROOTINO;
527 
528 			if ((error = read_inode(inumber, f)) != 0)
529 				goto out;
530 		}
531 	}
532 
533 	/*
534 	 * Found terminal component.
535 	 */
536 	error = 0;
537 	fp->f_seekp = 0;
538 out:
539 	if (buf)
540 		free(buf);
541 	if (path)
542 		free(path);
543 	if (error) {
544 		if (fp->f_buf)
545 			free(fp->f_buf);
546 		free(fp->f_fs);
547 		free(fp);
548 	}
549 	return (error);
550 }
551 
552 /*
553  * Read a new inode into a file structure.
554  */
555 static int
556 read_inode(ino_t inumber, struct open_file *f)
557 {
558 	struct file *fp = (struct file *)f->f_fsdata;
559 	struct ext2fs *fs = fp->f_fs;
560 	struct ext2dinode *dp;
561 	char *buf;
562 	size_t rsize;
563 	int level, error = 0;
564 
565 	/*
566 	 * Read inode and save it.
567 	 */
568 	buf = malloc(fs->fs_bsize);
569 	twiddle(1);
570 	error = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
571 	    ino_to_db(fs, fp->f_bg, inumber), fs->fs_bsize, buf, &rsize);
572 	if (error)
573 		goto out;
574 	if (rsize != fs->fs_bsize) {
575 		error = EIO;
576 		goto out;
577 	}
578 
579 	dp = (struct ext2dinode *)buf;
580 	fp->f_di = dp[ino_to_bo(fs, inumber)];
581 
582 	/* clear out old buffers */
583 	for (level = 0; level < EXT2_NIADDR; level++)
584 		fp->f_blkno[level] = -1;
585 	fp->f_buf_blkno = -1;
586 	fp->f_seekp = 0;
587 
588 out:
589 	free(buf);
590 	return (error);
591 }
592 
593 /*
594  * Given an offset in a file, find the disk block number that
595  * contains that block.
596  */
597 static int
598 block_map(struct open_file *f, daddr_t file_block, daddr_t *disk_block_p)
599 {
600 	struct file *fp = (struct file *)f->f_fsdata;
601 	struct ext2fs *fs = fp->f_fs;
602 	daddr_t ind_block_num;
603 	int32_t *ind_p;
604 	int idx, level;
605 	int error;
606 
607 	/*
608 	 * Index structure of an inode:
609 	 *
610 	 * di_db[0..EXT2_NDADDR-1] hold block numbers for blocks
611 	 *			0..EXT2_NDADDR-1
612 	 *
613 	 * di_ib[0]		index block 0 is the single indirect block
614 	 *			holds block numbers for blocks
615 	 *			EXT2_NDADDR .. EXT2_NDADDR + NINDIR(fs)-1
616 	 *
617 	 * di_ib[1]		index block 1 is the double indirect block
618 	 *			holds block numbers for INDEX blocks for blocks
619 	 *			EXT2_NDADDR + NINDIR(fs) ..
620 	 *			EXT2_NDADDR + NINDIR(fs) + NINDIR(fs)**2 - 1
621 	 *
622 	 * di_ib[2]		index block 2 is the triple indirect block
623 	 *			holds block numbers for double-indirect
624 	 *			blocks for blocks
625 	 *			EXT2_NDADDR + NINDIR(fs) + NINDIR(fs)**2 ..
626 	 *			EXT2_NDADDR + NINDIR(fs) + NINDIR(fs)**2
627 	 *				+ NINDIR(fs)**3 - 1
628 	 */
629 
630 	if (file_block < EXT2_NDADDR) {
631 		/* Direct block. */
632 		*disk_block_p = fp->f_di.di_db[file_block];
633 		return (0);
634 	}
635 
636 	file_block -= EXT2_NDADDR;
637 
638 	/*
639 	 * nindir[0] = NINDIR
640 	 * nindir[1] = NINDIR**2
641 	 * nindir[2] = NINDIR**3
642 	 *	etc
643 	 */
644 	for (level = 0; level < EXT2_NIADDR; level++) {
645 		if (file_block < fp->f_nindir[level])
646 			break;
647 		file_block -= fp->f_nindir[level];
648 	}
649 	if (level == EXT2_NIADDR) {
650 		/* Block number too high */
651 		return (EFBIG);
652 	}
653 
654 	ind_block_num = fp->f_di.di_ib[level];
655 
656 	for (; level >= 0; level--) {
657 		if (ind_block_num == 0) {
658 			*disk_block_p = 0;	/* missing */
659 			return (0);
660 		}
661 
662 		if (fp->f_blkno[level] != ind_block_num) {
663 			if (fp->f_blk[level] == (char *)0)
664 				fp->f_blk[level] =
665 					malloc(fs->fs_bsize);
666 			twiddle(1);
667 			error = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
668 			    fsb_to_db(fp->f_fs, ind_block_num), fs->fs_bsize,
669 			    fp->f_blk[level], &fp->f_blksize[level]);
670 			if (error)
671 				return (error);
672 			if (fp->f_blksize[level] != fs->fs_bsize)
673 				return (EIO);
674 			fp->f_blkno[level] = ind_block_num;
675 		}
676 
677 		ind_p = (int32_t *)fp->f_blk[level];
678 
679 		if (level > 0) {
680 			idx = file_block / fp->f_nindir[level - 1];
681 			file_block %= fp->f_nindir[level - 1];
682 		} else {
683 			idx = file_block;
684 		}
685 		ind_block_num = ind_p[idx];
686 	}
687 
688 	*disk_block_p = ind_block_num;
689 
690 	return (0);
691 }
692 
693 /*
694  * Read a portion of a file into an internal buffer.  Return
695  * the location in the buffer and the amount in the buffer.
696  */
697 static int
698 buf_read_file(struct open_file *f, char **buf_p, size_t *size_p)
699 {
700 	struct file *fp = (struct file *)f->f_fsdata;
701 	struct ext2fs *fs = fp->f_fs;
702 	long off;
703 	daddr_t file_block;
704 	daddr_t	disk_block;
705 	size_t block_size;
706 	int error = 0;
707 
708 	off = blkoff(fs, fp->f_seekp);
709 	file_block = lblkno(fs, fp->f_seekp);
710 	block_size = dblksize(fs, &fp->f_di, file_block);
711 
712 	if (file_block != fp->f_buf_blkno) {
713 		error = block_map(f, file_block, &disk_block);
714 		if (error)
715 			goto done;
716 
717 		if (fp->f_buf == (char *)0)
718 			fp->f_buf = malloc(fs->fs_bsize);
719 
720 		if (disk_block == 0) {
721 			bzero(fp->f_buf, block_size);
722 			fp->f_buf_size = block_size;
723 		} else {
724 			twiddle(4);
725 			error = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
726 			    fsb_to_db(fs, disk_block), block_size,
727 			    fp->f_buf, &fp->f_buf_size);
728 			if (error)
729 				goto done;
730 		}
731 		fp->f_buf_blkno = file_block;
732 	}
733 
734 	/*
735 	 * Return address of byte in buffer corresponding to
736 	 * offset, and size of remainder of buffer after that
737 	 * byte.
738 	 */
739 	*buf_p = fp->f_buf + off;
740 	*size_p = block_size - off;
741 
742 	/*
743 	 * But truncate buffer at end of file.
744 	 */
745 	if (*size_p > fp->f_di.di_size - fp->f_seekp)
746 		*size_p = fp->f_di.di_size - fp->f_seekp;
747 done:
748 	return (error);
749 }
750 
751 /*
752  * Search a directory for a name and return its
753  * i_number.
754  */
755 static int
756 search_directory(char *name, struct open_file *f, ino_t *inumber_p)
757 {
758 	struct file *fp = (struct file *)f->f_fsdata;
759 	struct ext2dirent *dp, *edp;
760 	char *buf;
761 	size_t buf_size;
762 	int namlen, length;
763 	int error;
764 
765 	length = strlen(name);
766 	fp->f_seekp = 0;
767 	while (fp->f_seekp < fp->f_di.di_size) {
768 		error = buf_read_file(f, &buf, &buf_size);
769 		if (error)
770 			return (error);
771 		dp = (struct ext2dirent *)buf;
772 		edp = (struct ext2dirent *)(buf + buf_size);
773 		while (dp < edp) {
774 			if (dp->d_ino == (ino_t)0)
775 				goto next;
776 			namlen = dp->d_namlen;
777 			if (namlen == length &&
778 			    strncmp(name, dp->d_name, length) == 0) {
779 				/* found entry */
780 				*inumber_p = dp->d_ino;
781 				return (0);
782 			}
783 		next:
784 			dp = (struct ext2dirent *)((char *)dp + dp->d_reclen);
785 		}
786 		fp->f_seekp += buf_size;
787 	}
788 	return (ENOENT);
789 }
790 
791 static int
792 ext2fs_close(struct open_file *f)
793 {
794 	struct file *fp = (struct file *)f->f_fsdata;
795 	int level;
796 
797 	if (fp == NULL)
798 		return (0);
799 
800 	f->f_fsdata = NULL;
801 
802 	for (level = 0; level < EXT2_NIADDR; level++) {
803 		if (fp->f_blk[level])
804 			free(fp->f_blk[level]);
805 	}
806 	if (fp->f_buf)
807 		free(fp->f_buf);
808 	if (fp->f_bg)
809 		free(fp->f_bg);
810 	free(fp->f_fs);
811 	free(fp);
812 	return (0);
813 }
814 
815 static int
816 ext2fs_read(struct open_file *f, void *addr, size_t size, size_t *resid)
817 {
818 	struct file *fp = (struct file *)f->f_fsdata;
819 	size_t csize, buf_size;
820 	char *buf;
821 	int error = 0;
822 
823 	while (size != 0) {
824 		if (fp->f_seekp >= fp->f_di.di_size)
825 			break;
826 
827 		error = buf_read_file(f, &buf, &buf_size);
828 		if (error)
829 			break;
830 
831 		csize = size;
832 		if (csize > buf_size)
833 			csize = buf_size;
834 
835 		bcopy(buf, addr, csize);
836 
837 		fp->f_seekp += csize;
838 		addr = (char *)addr + csize;
839 		size -= csize;
840 	}
841 	if (resid)
842 		*resid = size;
843 	return (error);
844 }
845 
846 static off_t
847 ext2fs_seek(struct open_file *f, off_t offset, int where)
848 {
849 	struct file *fp = (struct file *)f->f_fsdata;
850 
851 	switch (where) {
852 	case SEEK_SET:
853 		fp->f_seekp = offset;
854 		break;
855 	case SEEK_CUR:
856 		fp->f_seekp += offset;
857 		break;
858 	case SEEK_END:
859 		fp->f_seekp = fp->f_di.di_size - offset;
860 		break;
861 	default:
862 		errno = EINVAL;
863 		return (-1);
864 	}
865 	return (fp->f_seekp);
866 }
867 
868 static int
869 ext2fs_stat(struct open_file *f, struct stat *sb)
870 {
871 	struct file *fp = (struct file *)f->f_fsdata;
872 
873 	/* only important stuff */
874 	sb->st_mode = fp->f_di.di_mode;
875 	sb->st_uid = fp->f_di.di_uid;
876 	sb->st_gid = fp->f_di.di_gid;
877 	sb->st_size = fp->f_di.di_size;
878 	return (0);
879 }
880 
881 static int
882 ext2fs_readdir(struct open_file *f, struct dirent *d)
883 {
884 	struct file *fp = (struct file *)f->f_fsdata;
885 	struct ext2dirent *ed;
886 	char *buf;
887 	size_t buf_size;
888 	int error;
889 
890 	/*
891 	 * assume that a directory entry will not be split across blocks
892 	 */
893 
894 	do {
895 		if (fp->f_seekp >= fp->f_di.di_size)
896 			return (ENOENT);
897 		error = buf_read_file(f, &buf, &buf_size);
898 		if (error)
899 			return (error);
900 		ed = (struct ext2dirent *)buf;
901 		fp->f_seekp += ed->d_reclen;
902 	} while (ed->d_ino == (ino_t)0);
903 
904 	d->d_type = EXTFTODT(ed->d_type);
905 	strncpy(d->d_name, ed->d_name, ed->d_namlen);
906 	d->d_name[ed->d_namlen] = '\0';
907 	return (0);
908 }
909