1 /* $NetBSD: ufs.c,v 1.53 2008/01/02 11:48:58 ad Exp $ */
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed to Berkeley by
8 * The Mach Operating System project at Carnegie-Mellon University.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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 * 3. 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.
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
35 * Copyright (c) 1990, 1991 Carnegie Mellon University
36 * All Rights Reserved.
40 * Permission to use, copy, modify and distribute this software and its
41 * documentation is hereby granted, provided that both the copyright
42 * notice and this permission notice appear in all copies of the
43 * software, derivative works or modified versions, and any portions
44 * thereof, and that both notices appear in supporting documentation.
46 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
47 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
48 * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
50 * Carnegie Mellon requests users of this software to return to
52 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
53 * School of Computer Science
54 * Carnegie Mellon University
55 * Pittsburgh PA 15213-3890
57 * any improvements or extensions that they make and grant Carnegie the
58 * rights to redistribute these changes.
62 * Stand-alone file reading package for UFS and LFS filesystems.
65 #include <sys/param.h>
67 #include <ufs/ufs/dinode.h>
68 #include <ufs/ufs/dir.h>
70 #include <sys/queue.h>
71 #include <sys/condvar.h>
72 #include <sys/mount.h> /* XXX for MNAMELEN */
73 #include <ufs/lfs/lfs.h>
75 #include <ufs/ffs/fs.h>
78 #include <lib/libkern/libkern.h>
90 /* If this file is compiled by itself, build ufs (aka ffsv1) support */
91 #if !defined(LIBSA_FFSv2) && !defined(LIBSA_LFS)
95 #if defined(LIBSA_FS_SINGLECOMPONENT) && !defined(LIBSA_NO_FS_SYMLINK)
96 #define LIBSA_NO_FS_SYMLINK
98 #if defined(COMPAT_UFS) && defined(LIBSA_NO_COMPAT_UFS)
104 * In-core LFS superblock. This exists only to placate the macros in lfs.h,
107 struct dlfs lfs_dlfs
;
109 #define fs_magic lfs_magic
110 #define fs_maxsymlinklen lfs_maxsymlinklen
112 #define FS_MAGIC LFS_MAGIC
113 #define SBLOCKSIZE LFS_SBPAD
114 #define SBLOCKOFFSET LFS_LABELPAD
116 /* NB ufs2 doesn't use the common suberblock code... */
117 #define FS_MAGIC FS_UFS1_MAGIC
118 #define SBLOCKOFFSET SBLOCK_UFS1
121 #if defined(LIBSA_NO_TWIDDLE)
126 #if defined(LIBSA_FFSv2)
127 #define cgstart(fc, c) cgstart_ufs2((fs), (c))
129 #define cgstart(fc, c) cgstart_ufs1((fs), (c))
133 #define ufs_dinode ufs1_dinode
136 #define indp_t int32_t
138 typedef uint32_t ino32_t
;
140 #define FSBTODB(fs, indp) fsbtodb(fs, indp)
144 * To avoid having a lot of filesystem-block sized buffers lurking (which
145 * could be 32k) we only keep a few entries of the indirect block map.
146 * With 8k blocks, 2^8 blocks is ~500k so we reread the indirect block
147 * ~13 times pulling in a 6M kernel.
148 * The cache size must be smaller than the smallest filesystem block,
149 * so LN2_IND_CACHE_SZ <= 9 (UFS2 and 4k blocks).
151 #define LN2_IND_CACHE_SZ 6
152 #define IND_CACHE_SZ (1 << LN2_IND_CACHE_SZ)
153 #define IND_CACHE_MASK (IND_CACHE_SZ - 1)
159 off_t f_seekp
; /* seek pointer */
160 struct fs
*f_fs
; /* pointer to super-block */
161 struct ufs_dinode f_di
; /* copy of on-disk inode */
162 uint f_nishift
; /* for blocks in indirect block */
163 indp_t f_ind_cache_block
;
164 indp_t f_ind_cache
[IND_CACHE_SZ
];
166 char *f_buf
; /* buffer for data block */
167 size_t f_buf_size
; /* size of data block */
168 daddr_t f_buf_blkno
; /* block number of data block */
171 static int read_inode(ino32_t
, struct open_file
*);
172 static int block_map(struct open_file
*, indp_t
, indp_t
*);
173 static int buf_read_file(struct open_file
*, char **, size_t *);
174 static int search_directory(const char *, int, struct open_file
*, ino32_t
*);
176 static void ffs_oldfscompat(struct fs
*);
179 static int ffs_find_superblock(struct open_file
*, struct fs
*);
184 * Find an inode's block. Look it up in the ifile. Whee!
187 find_inode_sector(ino32_t inumber
, struct open_file
*f
, daddr_t
*isp
)
189 struct file
*fp
= (struct file
*)f
->f_fsdata
;
190 struct fs
*fs
= fp
->f_fs
;
191 daddr_t ifileent_blkno
;
193 size_t buf_after_ent
;
196 rc
= read_inode(fs
->lfs_ifile
, f
);
201 (inumber
/ fs
->lfs_ifpb
) + fs
->lfs_cleansz
+ fs
->lfs_segtabsz
;
202 fp
->f_seekp
= (off_t
)ifileent_blkno
* fs
->fs_bsize
+
203 (inumber
% fs
->lfs_ifpb
) * sizeof (IFILE_Vx
);
204 rc
= buf_read_file(f
, &ent_in_buf
, &buf_after_ent
);
207 /* make sure something's not badly wrong, but don't panic. */
208 if (buf_after_ent
< sizeof (IFILE_Vx
))
211 *isp
= FSBTODB(fs
, ((IFILE_Vx
*)ent_in_buf
)->if_daddr
);
212 if (*isp
== LFS_UNUSED_DADDR
) /* again, something badly wrong */
219 * Read a new inode into a file structure.
222 read_inode(ino32_t inumber
, struct open_file
*f
)
224 struct file
*fp
= (struct file
*)f
->f_fsdata
;
225 struct fs
*fs
= fp
->f_fs
;
229 daddr_t inode_sector
;
231 struct ufs_dinode
*dip
;
236 if (inumber
== fs
->lfs_ifile
)
237 inode_sector
= FSBTODB(fs
, fs
->lfs_idaddr
);
238 else if ((rc
= find_inode_sector(inumber
, f
, &inode_sector
)) != 0)
241 inode_sector
= FSBTODB(fs
, ino_to_fsba(fs
, inumber
));
245 * Read inode and save it.
249 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
, F_READ
,
250 inode_sector
, fs
->fs_bsize
, buf
, &rsize
);
253 if (rsize
!= fs
->fs_bsize
)
258 dip
= (struct ufs_dinode
*)buf
+ (cnt
- 1);
259 for (; dip
->di_inumber
!= inumber
; --dip
) {
260 /* kernel code panics, but boot blocks which panic are Bad. */
266 fp
->f_di
= ((struct ufs_dinode
*)buf
)[ino_to_fsbo(fs
, inumber
)];
270 * Clear out the old buffers
272 fp
->f_ind_cache_block
= ~0;
273 fp
->f_buf_blkno
= -1;
278 * Given an offset in a file, find the disk block number that
279 * contains that block.
282 block_map(struct open_file
*f
, indp_t file_block
, indp_t
*disk_block_p
)
284 struct file
*fp
= (struct file
*)f
->f_fsdata
;
285 struct fs
*fs
= fp
->f_fs
;
288 indp_t ind_block_num
;
291 indp_t
*buf
= (void *)fp
->f_buf
;
294 * Index structure of an inode:
296 * di_db[0..NDADDR-1] hold block numbers for blocks
299 * di_ib[0] index block 0 is the single indirect block
300 * holds block numbers for blocks
301 * NDADDR .. NDADDR + NINDIR(fs)-1
303 * di_ib[1] index block 1 is the double indirect block
304 * holds block numbers for INDEX blocks for blocks
305 * NDADDR + NINDIR(fs) ..
306 * NDADDR + NINDIR(fs) + NINDIR(fs)**2 - 1
308 * di_ib[2] index block 2 is the triple indirect block
309 * holds block numbers for double-indirect
311 * NDADDR + NINDIR(fs) + NINDIR(fs)**2 ..
312 * NDADDR + NINDIR(fs) + NINDIR(fs)**2
313 * + NINDIR(fs)**3 - 1
316 if (file_block
< NDADDR
) {
318 *disk_block_p
= fp
->f_di
.di_db
[file_block
];
322 file_block
-= NDADDR
;
324 ind_cache
= file_block
>> LN2_IND_CACHE_SZ
;
325 if (ind_cache
== fp
->f_ind_cache_block
) {
326 *disk_block_p
= fp
->f_ind_cache
[file_block
& IND_CACHE_MASK
];
331 level
+= fp
->f_nishift
;
332 if (file_block
< (indp_t
)1 << level
)
334 if (level
> NIADDR
* fp
->f_nishift
)
335 /* Block number too high */
337 file_block
-= (indp_t
)1 << level
;
340 ind_block_num
= fp
->f_di
.di_ib
[level
/ fp
->f_nishift
- 1];
343 level
-= fp
->f_nishift
;
344 if (ind_block_num
== 0) {
345 *disk_block_p
= 0; /* missing */
351 * If we were feeling brave, we could work out the number
352 * of the disk sector and read a single disk sector instead
353 * of a filesystem block.
354 * However we don't do this very often anyway...
356 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
, F_READ
,
357 FSBTODB(fp
->f_fs
, ind_block_num
), fs
->fs_bsize
,
361 if (rsize
!= fs
->fs_bsize
)
363 ind_block_num
= buf
[file_block
>> level
];
366 file_block
&= (1 << level
) - 1;
369 /* Save the part of the block that contains this sector */
370 memcpy(fp
->f_ind_cache
, &buf
[file_block
& ~IND_CACHE_MASK
],
371 IND_CACHE_SZ
* sizeof fp
->f_ind_cache
[0]);
372 fp
->f_ind_cache_block
= ind_cache
;
374 *disk_block_p
= ind_block_num
;
380 * Read a portion of a file into an internal buffer.
381 * Return the location in the buffer and the amount in the buffer.
384 buf_read_file(struct open_file
*f
, char **buf_p
, size_t *size_p
)
386 struct file
*fp
= (struct file
*)f
->f_fsdata
;
387 struct fs
*fs
= fp
->f_fs
;
394 off
= blkoff(fs
, fp
->f_seekp
);
395 file_block
= lblkno(fs
, fp
->f_seekp
);
397 block_size
= dblksize(fs
, &fp
->f_di
, file_block
);
399 block_size
= sblksize(fs
, (int64_t)fp
->f_di
.di_size
, file_block
);
402 if (file_block
!= fp
->f_buf_blkno
) {
403 rc
= block_map(f
, file_block
, &disk_block
);
407 if (disk_block
== 0) {
408 memset(fp
->f_buf
, 0, block_size
);
409 fp
->f_buf_size
= block_size
;
412 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
, F_READ
,
413 FSBTODB(fs
, disk_block
),
414 block_size
, fp
->f_buf
, &fp
->f_buf_size
);
419 fp
->f_buf_blkno
= file_block
;
423 * Return address of byte in buffer corresponding to
424 * offset, and size of remainder of buffer after that
427 *buf_p
= fp
->f_buf
+ off
;
428 *size_p
= block_size
- off
;
431 * But truncate buffer at end of file.
433 if (*size_p
> fp
->f_di
.di_size
- fp
->f_seekp
)
434 *size_p
= fp
->f_di
.di_size
- fp
->f_seekp
;
440 * Search a directory for a name and return its
444 search_directory(const char *name
, int length
, struct open_file
*f
,
447 struct file
*fp
= (struct file
*)f
->f_fsdata
;
456 while (fp
->f_seekp
< (off_t
)fp
->f_di
.di_size
) {
457 rc
= buf_read_file(f
, &buf
, &buf_size
);
461 dp
= (struct direct
*)buf
;
462 edp
= (struct direct
*)(buf
+ buf_size
);
463 for (;dp
< edp
; dp
= (void *)((char *)dp
+ dp
->d_reclen
)) {
464 if (dp
->d_reclen
<= 0)
466 if (dp
->d_ino
== (ino32_t
)0)
468 #if BYTE_ORDER == LITTLE_ENDIAN
469 if (fp
->f_fs
->fs_maxsymlinklen
<= 0)
473 namlen
= dp
->d_namlen
;
474 if (namlen
== length
&&
475 !memcmp(name
, dp
->d_name
, length
)) {
477 *inumber_p
= dp
->d_ino
;
481 fp
->f_seekp
+= buf_size
;
488 daddr_t sblock_try
[] = SBLOCKSEARCH
;
491 ffs_find_superblock(struct open_file
*f
, struct fs
*fs
)
496 for (i
= 0; sblock_try
[i
] != -1; i
++) {
497 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
, F_READ
,
498 sblock_try
[i
] / DEV_BSIZE
, SBLOCKSIZE
, fs
, &buf_size
);
499 if (rc
!= 0 || buf_size
!= SBLOCKSIZE
)
501 if (fs
->fs_sblockloc
!= sblock_try
[i
])
502 /* an alternate superblock - try again */
504 if (fs
->fs_magic
== FS_UFS2_MAGIC
) {
517 ufs_open(const char *path
, struct open_file
*f
)
519 #ifndef LIBSA_FS_SINGLECOMPONENT
520 const char *cp
, *ncp
;
527 #ifndef LIBSA_NO_FS_SYMLINK
528 ino32_t parent_inumber
;
530 char namebuf
[MAXPATHLEN
+1];
534 /* allocate file system specific data structure */
535 fp
= alloc(sizeof(struct file
));
536 memset(fp
, 0, sizeof(struct file
));
537 f
->f_fsdata
= (void *)fp
;
539 /* allocate space and read super block */
540 fs
= alloc(SBLOCKSIZE
);
545 rc
= ffs_find_superblock(f
, fs
);
551 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
, F_READ
,
552 SBLOCKOFFSET
/ DEV_BSIZE
, SBLOCKSIZE
, fs
, &buf_size
);
555 if (buf_size
!= SBLOCKSIZE
||
557 fs
->lfs_version
!= REQUIRED_LFS_VERSION
||
559 fs
->fs_magic
!= FS_MAGIC
) {
564 #if defined(LIBSA_LFS) && REQUIRED_LFS_VERSION == 2
566 * XXX We should check the second superblock and use the eldest
567 * of the two. See comments near the top of lfs_mountfs()
568 * in sys/ufs/lfs/lfs_vfsops.c.
569 * This may need a LIBSA_LFS_SMALL check as well.
578 if (fs
->fs_bsize
> MAXBSIZE
||
579 (size_t)fs
->fs_bsize
< sizeof(struct fs
)) {
585 * Calculate indirect block levels.
592 * We note that the number of indirect blocks is always
593 * a power of 2. This lets us use shifts and masks instead
594 * of divide and remainder and avoinds pulling in the
595 * 64bit division routine into the boot code.
599 if (mult
& (mult
- 1)) {
600 /* Hummm was't a power of 2 */
605 for (ln2
= 0; mult
!= 1; ln2
++)
611 /* alloc a block sized buffer used for all fs transfers */
612 fp
->f_buf
= alloc(fs
->fs_bsize
);
614 if ((rc
= read_inode(inumber
, f
)) != 0)
617 #ifndef LIBSA_FS_SINGLECOMPONENT
622 * Remove extra separators
630 * Check that current node is a directory.
632 if ((fp
->f_di
.di_mode
& IFMT
) != IFDIR
) {
638 * Get next component of path name.
641 while ((c
= *cp
) != '\0' && c
!= '/')
645 * Look up component in current directory.
646 * Save directory inumber in case we find a
649 #ifndef LIBSA_NO_FS_SYMLINK
650 parent_inumber
= inumber
;
652 rc
= search_directory(ncp
, cp
- ncp
, f
, &inumber
);
657 * Open next component.
659 if ((rc
= read_inode(inumber
, f
)) != 0)
662 #ifndef LIBSA_NO_FS_SYMLINK
664 * Check for symbolic link.
666 if ((fp
->f_di
.di_mode
& IFMT
) == IFLNK
) {
667 int link_len
= fp
->f_di
.di_size
;
672 if (link_len
+ len
> MAXPATHLEN
||
673 ++nlinks
> MAXSYMLINKS
) {
678 memmove(&namebuf
[link_len
], cp
, len
+ 1);
680 if (link_len
< fs
->fs_maxsymlinklen
) {
681 memcpy(namebuf
, fp
->f_di
.di_db
, link_len
);
684 * Read file for symbolic link
690 rc
= block_map(f
, (indp_t
)0, &disk_block
);
695 rc
= DEV_STRATEGY(f
->f_dev
)(f
->f_devdata
,
696 F_READ
, FSBTODB(fs
, disk_block
),
697 fs
->fs_bsize
, buf
, &buf_size
);
701 memcpy(namebuf
, buf
, link_len
);
705 * If relative pathname, restart at parent directory.
706 * If absolute pathname, restart at root.
710 inumber
= parent_inumber
;
712 inumber
= (ino32_t
)ROOTINO
;
714 if ((rc
= read_inode(inumber
, f
)) != 0)
717 #endif /* !LIBSA_NO_FS_SYMLINK */
721 * Found terminal component.
725 #else /* !LIBSA_FS_SINGLECOMPONENT */
727 /* look up component in the current (root) directory */
728 rc
= search_directory(path
, strlen(path
), f
, &inumber
);
733 rc
= read_inode(inumber
, f
);
735 #endif /* !LIBSA_FS_SINGLECOMPONENT */
737 fp
->f_seekp
= 0; /* reset seek pointer */
754 ufs_close(struct open_file
*f
)
756 struct file
*fp
= (struct file
*)f
->f_fsdata
;
763 dealloc(fp
->f_buf
, fp
->f_fs
->fs_bsize
);
764 dealloc(fp
->f_fs
, SBLOCKSIZE
);
765 dealloc(fp
, sizeof(struct file
));
770 * Copy a portion of a file into kernel memory.
771 * Cross block boundaries when necessary.
774 ufs_read(struct open_file
*f
, void *start
, size_t size
, size_t *resid
)
776 struct file
*fp
= (struct file
*)f
->f_fsdata
;
784 if (fp
->f_seekp
>= (off_t
)fp
->f_di
.di_size
)
787 rc
= buf_read_file(f
, &buf
, &buf_size
);
792 if (csize
> buf_size
)
795 memcpy(addr
, buf
, csize
);
797 fp
->f_seekp
+= csize
;
809 #ifndef LIBSA_NO_FS_WRITE
811 ufs_write(struct open_file
*f
, void *start
, size_t size
, size_t *resid
)
816 #endif /* !LIBSA_NO_FS_WRITE */
818 #ifndef LIBSA_NO_FS_SEEK
820 ufs_seek(struct open_file
*f
, off_t offset
, int where
)
822 struct file
*fp
= (struct file
*)f
->f_fsdata
;
826 fp
->f_seekp
= offset
;
829 fp
->f_seekp
+= offset
;
832 fp
->f_seekp
= fp
->f_di
.di_size
- offset
;
839 #endif /* !LIBSA_NO_FS_SEEK */
842 ufs_stat(struct open_file
*f
, struct stat
*sb
)
844 struct file
*fp
= (struct file
*)f
->f_fsdata
;
846 /* only important stuff */
847 memset(sb
, 0, sizeof *sb
);
848 sb
->st_mode
= fp
->f_di
.di_mode
;
849 sb
->st_uid
= fp
->f_di
.di_uid
;
850 sb
->st_gid
= fp
->f_di
.di_gid
;
851 sb
->st_size
= fp
->f_di
.di_size
;
857 * Sanity checks for old file systems.
859 * XXX - goes away some day.
860 * Stripped of stuff libsa doesn't need.....
863 ffs_oldfscompat(struct fs
*fs
)
868 * Newer Solaris versions have a slightly incompatible
869 * superblock - so always calculate this values on the fly, which
870 * is good enough for libsa purposes
872 if (fs
->fs_magic
== FS_UFS1_MAGIC
873 #ifndef COMPAT_SOLARIS_UFS
874 && fs
->fs_old_inodefmt
< FS_44INODEFMT
877 fs
->fs_qbmask
= ~fs
->fs_bmask
;
878 fs
->fs_qfmask
= ~fs
->fs_fmask
;