master
1/* $NetBSD: lfs.h,v 1.208 2020/03/28 01:08:42 christos Exp $ */
2
3/* from NetBSD: dinode.h,v 1.25 2016/01/22 23:06:10 dholland Exp */
4/* from NetBSD: dir.h,v 1.25 2015/09/01 06:16:03 dholland Exp */
5
6/*-
7 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
8 * All rights reserved.
9 *
10 * This code is derived from software contributed to The NetBSD Foundation
11 * by Konrad E. Schroder <perseant@hhhh.org>.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
24 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
25 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
26 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 * POSSIBILITY OF SUCH DAMAGE.
33 */
34/*-
35 * Copyright (c) 1991, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 *
62 * @(#)lfs.h 8.9 (Berkeley) 5/8/95
63 */
64/*
65 * Copyright (c) 2002 Networks Associates Technology, Inc.
66 * All rights reserved.
67 *
68 * This software was developed for the FreeBSD Project by Marshall
69 * Kirk McKusick and Network Associates Laboratories, the Security
70 * Research Division of Network Associates, Inc. under DARPA/SPAWAR
71 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
72 * research program
73 *
74 * Copyright (c) 1982, 1989, 1993
75 * The Regents of the University of California. All rights reserved.
76 * (c) UNIX System Laboratories, Inc.
77 * All or some portions of this file are derived from material licensed
78 * to the University of California by American Telephone and Telegraph
79 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
80 * the permission of UNIX System Laboratories, Inc.
81 *
82 * Redistribution and use in source and binary forms, with or without
83 * modification, are permitted provided that the following conditions
84 * are met:
85 * 1. Redistributions of source code must retain the above copyright
86 * notice, this list of conditions and the following disclaimer.
87 * 2. Redistributions in binary form must reproduce the above copyright
88 * notice, this list of conditions and the following disclaimer in the
89 * documentation and/or other materials provided with the distribution.
90 * 3. Neither the name of the University nor the names of its contributors
91 * may be used to endorse or promote products derived from this software
92 * without specific prior written permission.
93 *
94 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
95 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
96 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
97 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
98 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
99 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
100 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
101 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
102 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
103 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
104 * SUCH DAMAGE.
105 *
106 * @(#)dinode.h 8.9 (Berkeley) 3/29/95
107 */
108/*
109 * Copyright (c) 1982, 1986, 1989, 1993
110 * The Regents of the University of California. All rights reserved.
111 * (c) UNIX System Laboratories, Inc.
112 * All or some portions of this file are derived from material licensed
113 * to the University of California by American Telephone and Telegraph
114 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
115 * the permission of UNIX System Laboratories, Inc.
116 *
117 * Redistribution and use in source and binary forms, with or without
118 * modification, are permitted provided that the following conditions
119 * are met:
120 * 1. Redistributions of source code must retain the above copyright
121 * notice, this list of conditions and the following disclaimer.
122 * 2. Redistributions in binary form must reproduce the above copyright
123 * notice, this list of conditions and the following disclaimer in the
124 * documentation and/or other materials provided with the distribution.
125 * 3. Neither the name of the University nor the names of its contributors
126 * may be used to endorse or promote products derived from this software
127 * without specific prior written permission.
128 *
129 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
130 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
131 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
132 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
133 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
134 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
135 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
136 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
137 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
138 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
139 * SUCH DAMAGE.
140 *
141 * @(#)dir.h 8.5 (Berkeley) 4/27/95
142 */
143
144/*
145 * NOTE: COORDINATE ON-DISK FORMAT CHANGES WITH THE FREEBSD PROJECT.
146 */
147
148#ifndef _UFS_LFS_LFS_H_
149#define _UFS_LFS_LFS_H_
150
151#if !defined(_KERNEL) && !defined(_STANDALONE)
152#include <stddef.h> /* for offsetof */
153#endif
154
155#include <sys/rwlock.h>
156#include <sys/mutex.h>
157#include <sys/queue.h>
158#include <sys/condvar.h>
159#include <sys/mount.h>
160#include <sys/pool.h>
161
162/*
163 * Compile-time options for LFS.
164 */
165#define LFS_IFIND_RETRIES 16
166#define LFS_LOGLENGTH 1024 /* size of debugging log */
167#define LFS_MAX_ACTIVE 10 /* Dirty segments before ckp forced */
168
169/*
170 * Fixed filesystem layout parameters
171 */
172#define LFS_LABELPAD 8192 /* LFS label size */
173#define LFS_SBPAD 8192 /* LFS superblock size */
174
175#define LFS_UNUSED_INUM 0 /* 0: out of band inode number */
176#define LFS_IFILE_INUM 1 /* 1: IFILE inode number */
177 /* 2: Root inode number */
178#define LFS_LOSTFOUNDINO 3 /* 3: lost+found inode number */
179#define LFS_FIRST_INUM 4 /* 4: first free inode number */
180
181/*
182 * The root inode is the root of the file system. Inode 0 can't be used for
183 * normal purposes and historically bad blocks were linked to inode 1, thus
184 * the root inode is 2. (Inode 1 is no longer used for this purpose, however
185 * numerous dump tapes make this assumption, so we are stuck with it).
186 */
187#define ULFS_ROOTINO ((ino_t)2)
188
189/*
190 * The Whiteout inode# is a dummy non-zero inode number which will
191 * never be allocated to a real file. It is used as a place holder
192 * in the directory entry which has been tagged as a LFS_DT_WHT entry.
193 * See the comments about ULFS_ROOTINO above.
194 */
195#define ULFS_WINO ((ino_t)1)
196
197
198#define LFS_V1_SUMMARY_SIZE 512 /* V1 fixed summary size */
199#define LFS_DFL_SUMMARY_SIZE 512 /* Default summary size */
200
201#define LFS_MAXNAMLEN 255 /* maximum name length in a dir */
202
203#define ULFS_NXADDR 2
204#define ULFS_NDADDR 12 /* Direct addresses in inode. */
205#define ULFS_NIADDR 3 /* Indirect addresses in inode. */
206
207/*
208 * Adjustable filesystem parameters
209 */
210#ifndef LFS_ATIME_IFILE
211# define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */
212#endif
213#define LFS_MARKV_MAXBLKCNT 65536 /* Max block count for lfs_markv() */
214
215/*
216 * Directories
217 */
218
219/*
220 * Directories in LFS are files; they use the same inode and block
221 * mapping structures that regular files do. The directory per se is
222 * manifested in the file contents: an unordered, unstructured
223 * sequence of variable-size directory entries.
224 *
225 * This format and structure is taken (via what was originally shared
226 * ufs-level code) from FFS. Each directory entry is a fixed header
227 * followed by a string, the total length padded to a 4-byte boundary.
228 * All strings include a null terminator; the maximum string length
229 * is LFS_MAXNAMLEN, which is 255.
230 *
231 * The directory entry header structure (struct lfs_dirheader) is just
232 * the header information. A complete entry is this plus a null-
233 * terminated name following it, plus some amount of padding. The
234 * length of the name (not including the null terminator) is given by
235 * the namlen field of the header; the complete record length,
236 * including the null terminator and padding, is given by the reclen
237 * field of the header. The record length is always 4-byte aligned.
238 * (Even on 64-bit volumes, the record length is only 4-byte aligned,
239 * not 8-byte.)
240 *
241 * Historically, FFS directories were/are organized into blocks of
242 * size DIRBLKSIZE that can be written atomically to disk at the
243 * hardware level. Directory entries are not allowed to cross the
244 * boundaries of these blocks. The resulting atomicity is important
245 * for the integrity of FFS volumes; however, for LFS it's irrelevant.
246 * All we have to care about is not writing out directories that
247 * confuse earlier ufs-based versions of the LFS code.
248 *
249 * This means [to be determined]. (XXX)
250 *
251 * As DIRBLKSIZE in its FFS sense is hardware-dependent, and file
252 * system images do from time to time move to different hardware, code
253 * that reads directories should be prepared to handle directories
254 * written in a context where DIRBLKSIZE was different (smaller or
255 * larger) than its current value. Note however that it is not
256 * sensible for DIRBLKSIZE to be larger than the volume fragment size,
257 * and not practically possible for it to be larger than the volume
258 * block size.
259 *
260 * Some further notes:
261 * - the LFS_DIRSIZ macro provides the minimum space needed to hold
262 * a directory entry.
263 * - any particular entry may be arbitrarily larger (which is why the
264 * header stores both the entry size and the name size) to pad out
265 * unused space.
266 * - historically the padding in an entry is not necessarily zeroed
267 * but may contain trash.
268 * - dp->d_reclen is the size of the entry. This is always 4-byte
269 * aligned.
270 * - dp->d_namlen is the length of the string, and should always be
271 * the same as strlen(dp->d_name).
272 * - in particular, space available in an entry is given by
273 * dp->d_reclen - LFS_DIRSIZ(dp), and all space available within a
274 * directory block is tucked away within an existing entry.
275 * - all space within a directory block is part of some entry.
276 * - therefore, inserting a new entry requires finding and
277 * splitting a suitable existing entry, and when entries are
278 * removed their space is merged into the entry ahead of them.
279 * - an empty/unused entry has d_ino set to 0. This normally only
280 * appears in the first entry in a block, as elsewhere the unused
281 * entry should have been merged into the one before it. However,
282 * fsck leaves such entries behind so they must be tolerated
283 * elsewhere.
284 * - a completely empty directory block has one entry whose
285 * d_reclen is DIRBLKSIZ and whose d_ino is 0.
286 *
287 * The "old directory format" referenced by the fs->lfs_isolddirfmt
288 * flag (and some other things) refers to when the type field was
289 * added to directory entries. This change was made to FFS in the 80s,
290 * well before LFS was first written; there should be no LFS volumes
291 * (and certainly no LFS v2-format volumes or LFS64 volumes) where the
292 * old format pertains. All of the related logic should probably be
293 * removed; however, it hasn't been yet, and we get to carry it around
294 * until we can be conclusively sure it isn't needed.
295 *
296 * In the "old directory format" there is no type field and the namlen
297 * field is correspondingly 16 bits wide. On big-endian volumes this
298 * has no effect: namlen cannot exceed 255, so the upper byte is
299 * always 0 and this reads back from the type field as LFS_DT_UNKNOWN.
300 * On little-endian volumes, the namlen field will always be 0 and
301 * the namlen value needs to be read out of the type field. (The type
302 * is always LFS_DT_UNKNOWN.) The directory accessor functions take
303 * care of this so nothing else needs to be aware of it.
304 *
305 * LFS_OLDDIRFMT and LFS_NEWDIRFMT are code numbers for the old and
306 * new directory format respectively. These codes do not appear on
307 * disk; they're generated from a runtime macro called FSFMT() that's
308 * cued by other things. This is why (confusingly) LFS_OLDDIRFMT is 1
309 * and LFS_NEWDIRFMT is 0.
310 *
311 * FSFMT(), LFS_OLDDIRFMT, and LFS_NEWDIRFMT should be removed. (XXX)
312 */
313
314/*
315 * Directory block size.
316 */
317#undef LFS_DIRBLKSIZ
318#define LFS_DIRBLKSIZ DEV_BSIZE
319
320/*
321 * Convert between stat structure type codes and directory entry type codes.
322 */
323#define LFS_IFTODT(mode) (((mode) & 0170000) >> 12)
324#define LFS_DTTOIF(dirtype) ((dirtype) << 12)
325
326/*
327 * Theoretically, directories can be more than 2Gb in length; however, in
328 * practice this seems unlikely. So, we define the type doff_t as a 32-bit
329 * quantity to keep down the cost of doing lookup on a 32-bit machine.
330 */
331#define doff_t int32_t
332#define lfs_doff_t int32_t
333#define LFS_MAXDIRSIZE (0x7fffffff)
334
335/*
336 * File types for d_type
337 */
338#define LFS_DT_UNKNOWN 0
339#define LFS_DT_FIFO 1
340#define LFS_DT_CHR 2
341#define LFS_DT_DIR 4
342#define LFS_DT_BLK 6
343#define LFS_DT_REG 8
344#define LFS_DT_LNK 10
345#define LFS_DT_SOCK 12
346#define LFS_DT_WHT 14
347
348/*
349 * (See notes above)
350 */
351
352struct lfs_dirheader32 {
353 uint32_t dh_ino; /* inode number of entry */
354 uint16_t dh_reclen; /* length of this record */
355 uint8_t dh_type; /* file type, see below */
356 uint8_t dh_namlen; /* length of string in d_name */
357};
358__CTASSERT(sizeof(struct lfs_dirheader32) == 8);
359
360struct lfs_dirheader64 {
361 uint64_t dh_ino; /* inode number of entry */
362 uint16_t dh_reclen; /* length of this record */
363 uint8_t dh_type; /* file type, see below */
364 uint8_t dh_namlen; /* length of string in d_name */
365} __aligned(4) __packed;
366__CTASSERT(sizeof(struct lfs_dirheader64) == 12);
367
368union lfs_dirheader {
369 struct lfs_dirheader64 u_64;
370 struct lfs_dirheader32 u_32;
371};
372__CTASSERT(__alignof(union lfs_dirheader) == __alignof(struct lfs_dirheader64));
373#ifndef __lint__
374__CTASSERT(__alignof(union lfs_dirheader) == __alignof(struct lfs_dirheader32));
375#endif
376
377typedef union lfs_dirheader LFS_DIRHEADER;
378
379/*
380 * Template for manipulating directories.
381 */
382
383struct lfs_dirtemplate32 {
384 struct lfs_dirheader32 dot_header;
385 char dot_name[4]; /* must be multiple of 4 */
386 struct lfs_dirheader32 dotdot_header;
387 char dotdot_name[4]; /* ditto */
388};
389__CTASSERT(sizeof(struct lfs_dirtemplate32) == 2*(8 + 4));
390
391struct lfs_dirtemplate64 {
392 struct lfs_dirheader64 dot_header;
393 char dot_name[4]; /* must be multiple of 4 */
394 struct lfs_dirheader64 dotdot_header;
395 char dotdot_name[4]; /* ditto */
396};
397__CTASSERT(sizeof(struct lfs_dirtemplate64) == 2*(12 + 4));
398
399union lfs_dirtemplate {
400 struct lfs_dirtemplate64 u_64;
401 struct lfs_dirtemplate32 u_32;
402};
403
404#if 0
405/*
406 * This is the old format of directories, sans type element.
407 */
408struct lfs_odirtemplate {
409 uint32_t dot_ino;
410 int16_t dot_reclen;
411 uint16_t dot_namlen;
412 char dot_name[4]; /* must be multiple of 4 */
413 uint32_t dotdot_ino;
414 int16_t dotdot_reclen;
415 uint16_t dotdot_namlen;
416 char dotdot_name[4]; /* ditto */
417};
418__CTASSERT(sizeof(struct lfs_odirtemplate) == 2*(8 + 4));
419#endif
420
421/*
422 * Inodes
423 */
424
425/*
426 * A dinode contains all the meta-data associated with a LFS file.
427 * This structure defines the on-disk format of a dinode. Since
428 * this structure describes an on-disk structure, all its fields
429 * are defined by types with precise widths.
430 */
431
432struct lfs32_dinode {
433 uint16_t di_mode; /* 0: IFMT, permissions; see below. */
434 int16_t di_nlink; /* 2: File link count. */
435 uint32_t di_inumber; /* 4: Inode number. */
436 uint64_t di_size; /* 8: File byte count. */
437 int32_t di_atime; /* 16: Last access time. */
438 int32_t di_atimensec; /* 20: Last access time. */
439 int32_t di_mtime; /* 24: Last modified time. */
440 int32_t di_mtimensec; /* 28: Last modified time. */
441 int32_t di_ctime; /* 32: Last inode change time. */
442 int32_t di_ctimensec; /* 36: Last inode change time. */
443 int32_t di_db[ULFS_NDADDR]; /* 40: Direct disk blocks. */
444 int32_t di_ib[ULFS_NIADDR]; /* 88: Indirect disk blocks. */
445 uint32_t di_flags; /* 100: Status flags (chflags). */
446 uint32_t di_blocks; /* 104: Blocks actually held. */
447 int32_t di_gen; /* 108: Generation number. */
448 uint32_t di_uid; /* 112: File owner. */
449 uint32_t di_gid; /* 116: File group. */
450 uint64_t di_modrev; /* 120: i_modrev for NFSv4 */
451};
452__CTASSERT(sizeof(struct lfs32_dinode) == 128);
453
454struct lfs64_dinode {
455 uint16_t di_mode; /* 0: IFMT, permissions; see below. */
456 int16_t di_nlink; /* 2: File link count. */
457 uint32_t di_uid; /* 4: File owner. */
458 uint32_t di_gid; /* 8: File group. */
459 uint32_t di_blksize; /* 12: Inode blocksize. */
460 uint64_t di_size; /* 16: File byte count. */
461 uint64_t di_blocks; /* 24: Bytes actually held. */
462 int64_t di_atime; /* 32: Last access time. */
463 int64_t di_mtime; /* 40: Last modified time. */
464 int64_t di_ctime; /* 48: Last inode change time. */
465 int64_t di_birthtime; /* 56: Inode creation time. */
466 int32_t di_mtimensec; /* 64: Last modified time. */
467 int32_t di_atimensec; /* 68: Last access time. */
468 int32_t di_ctimensec; /* 72: Last inode change time. */
469 int32_t di_birthnsec; /* 76: Inode creation time. */
470 int32_t di_gen; /* 80: Generation number. */
471 uint32_t di_kernflags; /* 84: Kernel flags. */
472 uint32_t di_flags; /* 88: Status flags (chflags). */
473 int32_t di_extsize; /* 92: External attributes block. */
474 int64_t di_extb[ULFS_NXADDR];/* 96: External attributes block. */
475 int64_t di_db[ULFS_NDADDR]; /* 112: Direct disk blocks. */
476 int64_t di_ib[ULFS_NIADDR]; /* 208: Indirect disk blocks. */
477 uint64_t di_modrev; /* 232: i_modrev for NFSv4 */
478 uint64_t di_inumber; /* 240: Inode number */
479 uint64_t di_spare[1]; /* 248: Reserved; currently unused */
480};
481__CTASSERT(sizeof(struct lfs64_dinode) == 256);
482
483union lfs_dinode {
484 struct lfs64_dinode u_64;
485 struct lfs32_dinode u_32;
486};
487__CTASSERT(__alignof(union lfs_dinode) == __alignof(struct lfs64_dinode));
488__CTASSERT(__alignof(union lfs_dinode) == __alignof(struct lfs32_dinode));
489
490/*
491 * The di_db fields may be overlaid with other information for
492 * file types that do not have associated disk storage. Block
493 * and character devices overlay the first data block with their
494 * dev_t value. Short symbolic links place their path in the
495 * di_db area.
496 */
497#define di_rdev di_db[0]
498
499/* Size of the on-disk inode. */
500//#define LFS_DINODE1_SIZE (sizeof(struct ulfs1_dinode)) /* 128 */
501//#define LFS_DINODE2_SIZE (sizeof(struct ulfs2_dinode))
502
503/* File types, found in the upper bits of di_mode. */
504#define LFS_IFMT 0170000 /* Mask of file type. */
505#define LFS_IFIFO 0010000 /* Named pipe (fifo). */
506#define LFS_IFCHR 0020000 /* Character device. */
507#define LFS_IFDIR 0040000 /* Directory file. */
508#define LFS_IFBLK 0060000 /* Block device. */
509#define LFS_IFREG 0100000 /* Regular file. */
510#define LFS_IFLNK 0120000 /* Symbolic link. */
511#define LFS_IFSOCK 0140000 /* UNIX domain socket. */
512#define LFS_IFWHT 0160000 /* Whiteout. */
513
514/*
515 * "struct buf" associated definitions
516 */
517
518/* Unassigned disk addresses. */
519#define UNASSIGNED -1
520#define UNWRITTEN -2
521
522/* Unused logical block number */
523#define LFS_UNUSED_LBN -1
524
525/*
526 * On-disk and in-memory checkpoint segment usage structure.
527 */
528typedef struct segusage SEGUSE;
529struct segusage {
530 uint32_t su_nbytes; /* 0: number of live bytes */
531 uint32_t su_olastmod; /* 4: SEGUSE last modified timestamp */
532 uint16_t su_nsums; /* 8: number of summaries in segment */
533 uint16_t su_ninos; /* 10: number of inode blocks in seg */
534
535#define SEGUSE_ACTIVE 0x01 /* segment currently being written */
536#define SEGUSE_DIRTY 0x02 /* segment has data in it */
537#define SEGUSE_SUPERBLOCK 0x04 /* segment contains a superblock */
538#define SEGUSE_ERROR 0x08 /* cleaner: do not clean segment */
539#define SEGUSE_EMPTY 0x10 /* segment is empty */
540#define SEGUSE_INVAL 0x20 /* segment is invalid */
541 uint32_t su_flags; /* 12: segment flags */
542 uint64_t su_lastmod; /* 16: last modified timestamp */
543};
544__CTASSERT(sizeof(struct segusage) == 24);
545
546typedef struct segusage_v1 SEGUSE_V1;
547struct segusage_v1 {
548 uint32_t su_nbytes; /* 0: number of live bytes */
549 uint32_t su_lastmod; /* 4: SEGUSE last modified timestamp */
550 uint16_t su_nsums; /* 8: number of summaries in segment */
551 uint16_t su_ninos; /* 10: number of inode blocks in seg */
552 uint32_t su_flags; /* 12: segment flags */
553};
554__CTASSERT(sizeof(struct segusage_v1) == 16);
555
556/*
557 * On-disk file information. One per file with data blocks in the segment.
558 *
559 * The FINFO structure is a header; it is followed by fi_nblocks block
560 * pointers, which are logical block numbers of the file. (These are the
561 * blocks of the file present in this segment.)
562 */
563
564typedef struct finfo64 FINFO64;
565struct finfo64 {
566 uint32_t fi_nblocks; /* number of blocks */
567 uint32_t fi_version; /* version number */
568 uint64_t fi_ino; /* inode number */
569 uint32_t fi_lastlength; /* length of last block in array */
570 uint32_t fi_pad; /* unused */
571} __aligned(4) __packed;
572__CTASSERT(sizeof(struct finfo64) == 24);
573
574typedef struct finfo32 FINFO32;
575struct finfo32 {
576 uint32_t fi_nblocks; /* number of blocks */
577 uint32_t fi_version; /* version number */
578 uint32_t fi_ino; /* inode number */
579 uint32_t fi_lastlength; /* length of last block in array */
580};
581__CTASSERT(sizeof(struct finfo32) == 16);
582
583typedef union finfo {
584 struct finfo64 u_64;
585 struct finfo32 u_32;
586} FINFO;
587__CTASSERT(__alignof(union finfo) == __alignof(struct finfo64));
588#ifndef __lint__
589__CTASSERT(__alignof(union finfo) == __alignof(struct finfo32));
590#endif
591
592/*
593 * inode info (part of the segment summary)
594 *
595 * Each one of these is just a block number; wrapping the structure
596 * around it gives more contextual information in the code about
597 * what's going on.
598 */
599
600typedef struct iinfo64 {
601 uint64_t ii_block; /* block number */
602} __aligned(4) __packed IINFO64;
603__CTASSERT(sizeof(struct iinfo64) == 8);
604
605typedef struct iinfo32 {
606 uint32_t ii_block; /* block number */
607} IINFO32;
608__CTASSERT(sizeof(struct iinfo32) == 4);
609
610typedef union iinfo {
611 struct iinfo64 u_64;
612 struct iinfo32 u_32;
613} IINFO;
614__CTASSERT(__alignof(union iinfo) == __alignof(struct iinfo64));
615#ifndef __lint__
616__CTASSERT(__alignof(union iinfo) == __alignof(struct iinfo32));
617#endif
618
619/*
620 * Index file inode entries.
621 */
622
623/* magic value for daddrs */
624#define LFS_UNUSED_DADDR 0 /* out-of-band daddr */
625/* magic value for if_nextfree -- indicate orphaned file */
626#define LFS_ORPHAN_NEXTFREE(fs) \
627 ((fs)->lfs_is64 ? ~(uint64_t)0 : ~(uint32_t)0)
628
629typedef struct ifile64 IFILE64;
630struct ifile64 {
631 uint32_t if_version; /* inode version number */
632 uint32_t if_atime_nsec; /* and nanoseconds */
633 uint64_t if_atime_sec; /* Last access time, seconds */
634 int64_t if_daddr; /* inode disk address */
635 uint64_t if_nextfree; /* next-unallocated inode */
636} __aligned(4) __packed;
637__CTASSERT(sizeof(struct ifile64) == 32);
638
639typedef struct ifile32 IFILE32;
640struct ifile32 {
641 uint32_t if_version; /* inode version number */
642 int32_t if_daddr; /* inode disk address */
643 uint32_t if_nextfree; /* next-unallocated inode */
644 uint32_t if_atime_sec; /* Last access time, seconds */
645 uint32_t if_atime_nsec; /* and nanoseconds */
646};
647__CTASSERT(sizeof(struct ifile32) == 20);
648
649typedef struct ifile_v1 IFILE_V1;
650struct ifile_v1 {
651 uint32_t if_version; /* inode version number */
652 int32_t if_daddr; /* inode disk address */
653 uint32_t if_nextfree; /* next-unallocated inode */
654#if LFS_ATIME_IFILE
655#error "this cannot work"
656 struct timespec if_atime; /* Last access time */
657#endif
658};
659__CTASSERT(sizeof(struct ifile_v1) == 12);
660
661/*
662 * Note: struct ifile_v1 is often handled by accessing the first three
663 * fields of struct ifile32. (XXX: Blah. This should be cleaned up as
664 * it may in some cases violate the strict-aliasing rules.)
665 */
666typedef union ifile {
667 struct ifile64 u_64;
668 struct ifile32 u_32;
669 struct ifile_v1 u_v1;
670} IFILE;
671__CTASSERT(__alignof(union ifile) == __alignof(struct ifile64));
672#ifndef __lint__
673__CTASSERT(__alignof(union ifile) == __alignof(struct ifile32));
674__CTASSERT(__alignof(union ifile) == __alignof(struct ifile_v1));
675#endif
676
677/*
678 * Cleaner information structure. This resides in the ifile and is used
679 * to pass information from the kernel to the cleaner.
680 */
681
682/* flags for ->flags */
683#define LFS_CLEANER_MUST_CLEAN 0x01
684
685typedef struct _cleanerinfo32 {
686 uint32_t clean; /* 0: number of clean segments */
687 uint32_t dirty; /* 4: number of dirty segments */
688 int32_t bfree; /* 8: disk blocks free */
689 int32_t avail; /* 12: disk blocks available */
690 uint32_t free_head; /* 16: head of the inode free list */
691 uint32_t free_tail; /* 20: tail of the inode free list */
692 uint32_t flags; /* 24: status word from the kernel */
693} CLEANERINFO32;
694__CTASSERT(sizeof(struct _cleanerinfo32) == 28);
695
696typedef struct _cleanerinfo64 {
697 uint32_t clean; /* 0: number of clean segments */
698 uint32_t dirty; /* 4: number of dirty segments */
699 int64_t bfree; /* 8: disk blocks free */
700 int64_t avail; /* 16: disk blocks available */
701 uint64_t free_head; /* 24: head of the inode free list */
702 uint64_t free_tail; /* 32: tail of the inode free list */
703 uint32_t flags; /* 40: status word from the kernel */
704 uint32_t pad; /* 44: must be 64-bit aligned */
705} __aligned(4) __packed CLEANERINFO64;
706__CTASSERT(sizeof(struct _cleanerinfo64) == 48);
707
708/* this must not go to disk directly of course */
709typedef union _cleanerinfo {
710 CLEANERINFO32 u_32;
711 CLEANERINFO64 u_64;
712} CLEANERINFO;
713#ifndef __lint__
714__CTASSERT(__alignof(union _cleanerinfo) == __alignof(struct _cleanerinfo32));
715__CTASSERT(__alignof(union _cleanerinfo) == __alignof(struct _cleanerinfo64));
716#endif
717
718/*
719 * On-disk segment summary information
720 */
721
722/* magic value for ss_magic */
723#define SS_MAGIC 0x061561
724
725/* flags for ss_flags */
726#define SS_DIROP 0x01 /* segment begins a dirop */
727#define SS_CONT 0x02 /* more partials to finish this write*/
728#define SS_CLEAN 0x04 /* written by the cleaner */
729#define SS_RFW 0x08 /* written by the roll-forward agent */
730#define SS_RECLAIM 0x10 /* written by the roll-forward agent */
731
732/* type used for reading checksum signatures from metadata structures */
733typedef uint32_t lfs_checkword;
734
735typedef struct segsum_v1 SEGSUM_V1;
736struct segsum_v1 {
737 uint32_t ss_sumsum; /* 0: check sum of summary block */
738 uint32_t ss_datasum; /* 4: check sum of data */
739 uint32_t ss_magic; /* 8: segment summary magic number */
740 int32_t ss_next; /* 12: next segment */
741 uint32_t ss_create; /* 16: creation time stamp */
742 uint16_t ss_nfinfo; /* 20: number of file info structures */
743 uint16_t ss_ninos; /* 22: number of inodes in summary */
744 uint16_t ss_flags; /* 24: used for directory operations */
745 uint16_t ss_pad; /* 26: extra space */
746 /* FINFO's and inode daddr's... */
747};
748__CTASSERT(sizeof(struct segsum_v1) == 28);
749
750typedef struct segsum32 SEGSUM32;
751struct segsum32 {
752 uint32_t ss_sumsum; /* 0: check sum of summary block */
753 uint32_t ss_datasum; /* 4: check sum of data */
754 uint32_t ss_magic; /* 8: segment summary magic number */
755 int32_t ss_next; /* 12: next segment (disk address) */
756 uint32_t ss_ident; /* 16: roll-forward fsid */
757 uint16_t ss_nfinfo; /* 20: number of file info structures */
758 uint16_t ss_ninos; /* 22: number of inodes in summary */
759 uint16_t ss_flags; /* 24: used for directory operations */
760 uint8_t ss_pad[2]; /* 26: extra space */
761 uint32_t ss_reclino; /* 28: inode being reclaimed */
762 uint64_t ss_serial; /* 32: serial number */
763 uint64_t ss_create; /* 40: time stamp */
764 /* FINFO's and inode daddr's... */
765} __aligned(4) __packed;
766__CTASSERT(sizeof(struct segsum32) == 48);
767
768typedef struct segsum64 SEGSUM64;
769struct segsum64 {
770 uint32_t ss_sumsum; /* 0: check sum of summary block */
771 uint32_t ss_datasum; /* 4: check sum of data */
772 uint32_t ss_magic; /* 8: segment summary magic number */
773 uint32_t ss_ident; /* 12: roll-forward fsid */
774 int64_t ss_next; /* 16: next segment (disk address) */
775 uint16_t ss_nfinfo; /* 24: number of file info structures */
776 uint16_t ss_ninos; /* 26: number of inodes in summary */
777 uint16_t ss_flags; /* 28: used for directory operations */
778 uint8_t ss_pad[2]; /* 30: extra space */
779 uint64_t ss_reclino; /* 32: inode being reclaimed */
780 uint64_t ss_serial; /* 40: serial number */
781 uint64_t ss_create; /* 48: time stamp */
782 /* FINFO's and inode daddr's... */
783} __aligned(4) __packed;
784__CTASSERT(sizeof(struct segsum64) == 56);
785
786typedef union segsum SEGSUM;
787union segsum {
788 struct segsum64 u_64;
789 struct segsum32 u_32;
790 struct segsum_v1 u_v1;
791};
792__CTASSERT(__alignof(union segsum) == __alignof(struct segsum64));
793__CTASSERT(__alignof(union segsum) == __alignof(struct segsum32));
794#ifndef __lint__
795__CTASSERT(__alignof(union segsum) == __alignof(struct segsum_v1));
796#endif
797
798/*
799 * On-disk super block.
800 *
801 * We have separate superblock structures for the 32-bit and 64-bit
802 * LFS, and accessor functions to hide the differences.
803 *
804 * For lfs64, the format version is always 2; version 1 lfs is old.
805 * For both, the inode format version is 0; for lfs32 this selects the
806 * same 32-bit inode as always, and for lfs64 this selects the larger
807 * 64-bit inode structure we got from ffsv2.
808 *
809 * In lfs64:
810 * - inode numbers are 64 bit now
811 * - segments may not be larger than 4G (counted in bytes)
812 * - there may not be more than 2^32 (or perhaps 2^31) segments
813 * - the total volume size is limited to 2^63 frags and/or 2^63
814 * disk blocks, and probably in practice 2^63 bytes.
815 */
816
817#define LFS_MAGIC 0x070162
818#define LFS_MAGIC_SWAPPED 0x62010700
819
820#define LFS64_MAGIC (0x19620701 ^ 0xffffffff)
821#define LFS64_MAGIC_SWAPPED (0x01076219 ^ 0xffffffff)
822
823#define LFS_VERSION 2
824
825#define LFS_MIN_SBINTERVAL 5 /* min superblock segment spacing */
826#define LFS_MAXNUMSB 10 /* max number of superblocks */
827
828/* flags for dlfs_pflags */
829#define LFS_PF_CLEAN 0x1
830
831/* Inode format versions */
832#define LFS_44INODEFMT 0
833#define LFS_MAXINODEFMT 0
834
835struct dlfs {
836 uint32_t dlfs_magic; /* 0: magic number */
837 uint32_t dlfs_version; /* 4: version number */
838
839 uint32_t dlfs_size; /* 8: number of blocks in fs (v1) */
840 /* number of frags in fs (v2) */
841 uint32_t dlfs_ssize; /* 12: number of blocks per segment (v1) */
842 /* number of bytes per segment (v2) */
843 uint32_t dlfs_dsize; /* 16: number of disk blocks in fs */
844 uint32_t dlfs_bsize; /* 20: file system block size */
845 uint32_t dlfs_fsize; /* 24: size of frag blocks in fs */
846 uint32_t dlfs_frag; /* 28: number of frags in a block in fs */
847
848/* Checkpoint region. */
849 uint32_t dlfs_freehd; /* 32: start of the free inode list */
850 int32_t dlfs_bfree; /* 36: number of free frags */
851 uint32_t dlfs_nfiles; /* 40: number of allocated inodes */
852 int32_t dlfs_avail; /* 44: blocks available for writing */
853 int32_t dlfs_uinodes; /* 48: inodes in cache not yet on disk */
854 int32_t dlfs_idaddr; /* 52: inode file disk address */
855 uint32_t dlfs_ifile; /* 56: inode file inode number */
856 int32_t dlfs_lastseg; /* 60: address of last segment written */
857 int32_t dlfs_nextseg; /* 64: address of next segment to write */
858 int32_t dlfs_curseg; /* 68: current segment being written */
859 int32_t dlfs_offset; /* 72: offset in curseg for next partial */
860 int32_t dlfs_lastpseg; /* 76: address of last partial written */
861 uint32_t dlfs_inopf; /* 80: v1: time stamp; v2: inodes per frag */
862
863/* These are configuration parameters. */
864 uint32_t dlfs_minfree; /* 84: minimum percentage of free blocks */
865
866/* These fields can be computed from the others. */
867 uint64_t dlfs_maxfilesize; /* 88: maximum representable file size */
868 uint32_t dlfs_fsbpseg; /* 96: frags (fsb) per segment */
869 uint32_t dlfs_inopb; /* 100: inodes per block */
870 uint32_t dlfs_ifpb; /* 104: IFILE entries per block */
871 uint32_t dlfs_sepb; /* 108: SEGUSE entries per block */
872 uint32_t dlfs_nindir; /* 112: indirect pointers per block */
873 uint32_t dlfs_nseg; /* 116: number of segments */
874 uint32_t dlfs_nspf; /* 120: number of sectors per fragment */
875 uint32_t dlfs_cleansz; /* 124: cleaner info size in blocks */
876 uint32_t dlfs_segtabsz; /* 128: segment table size in blocks */
877 uint32_t dlfs_segmask; /* 132: calculate offset within a segment */
878 uint32_t dlfs_segshift; /* 136: fast mult/div for segments */
879 uint32_t dlfs_bshift; /* 140: calc block number from file offset */
880 uint32_t dlfs_ffshift; /* 144: fast mult/div for frag from file */
881 uint32_t dlfs_fbshift; /* 148: fast mult/div for frag from block */
882 uint64_t dlfs_bmask; /* 152: calc block offset from file offset */
883 uint64_t dlfs_ffmask; /* 160: calc frag offset from file offset */
884 uint64_t dlfs_fbmask; /* 168: calc frag offset from block offset */
885 uint32_t dlfs_blktodb; /* 176: blktodb and dbtoblk shift constant */
886 uint32_t dlfs_sushift; /* 180: fast mult/div for segusage table */
887
888 int32_t dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
889 /* 188: superblock disk offsets */
890 int32_t dlfs_sboffs[LFS_MAXNUMSB];
891
892 uint32_t dlfs_nclean; /* 228: Number of clean segments */
893 u_char dlfs_fsmnt[MNAMELEN]; /* 232: name mounted on */
894 uint16_t dlfs_pflags; /* 322: file system persistent flags */
895 int32_t dlfs_dmeta; /* 324: total number of dirty summaries */
896 uint32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */
897 uint32_t dlfs_sumsize; /* 332: size of summary blocks */
898 uint64_t dlfs_serial; /* 336: serial number */
899 uint32_t dlfs_ibsize; /* 344: size of inode blocks */
900 int32_t dlfs_s0addr; /* 348: start of segment 0 */
901 uint64_t dlfs_tstamp; /* 352: time stamp */
902 uint32_t dlfs_inodefmt; /* 360: inode format version */
903 uint32_t dlfs_interleave; /* 364: segment interleave */
904 uint32_t dlfs_ident; /* 368: per-fs identifier */
905 uint32_t dlfs_fsbtodb; /* 372: fsbtodb and dbtodsb shift constant */
906 uint32_t dlfs_resvseg; /* 376: segments reserved for the cleaner */
907 int8_t dlfs_pad[128]; /* 380: round to 512 bytes */
908/* Checksum -- last valid disk field. */
909 uint32_t dlfs_cksum; /* 508: checksum for superblock checking */
910};
911
912struct dlfs64 {
913 uint32_t dlfs_magic; /* 0: magic number */
914 uint32_t dlfs_version; /* 4: version number (2) */
915
916 uint64_t dlfs_size; /* 8: number of frags in fs (v2) */
917 uint64_t dlfs_dsize; /* 16: number of disk blocks in fs */
918 uint32_t dlfs_ssize; /* 24: number of bytes per segment (v2) */
919 uint32_t dlfs_bsize; /* 28: file system block size */
920 uint32_t dlfs_fsize; /* 32: size of frag blocks in fs */
921 uint32_t dlfs_frag; /* 36: number of frags in a block in fs */
922
923/* Checkpoint region. */
924 uint64_t dlfs_freehd; /* 40: start of the free inode list */
925 uint64_t dlfs_nfiles; /* 48: number of allocated inodes */
926 int64_t dlfs_bfree; /* 56: number of free frags */
927 int64_t dlfs_avail; /* 64: blocks available for writing */
928 int64_t dlfs_idaddr; /* 72: inode file disk address */
929 int32_t dlfs_uinodes; /* 80: inodes in cache not yet on disk */
930 uint32_t dlfs_unused_0; /* 84: not used */
931 int64_t dlfs_lastseg; /* 88: address of last segment written */
932 int64_t dlfs_nextseg; /* 96: address of next segment to write */
933 int64_t dlfs_curseg; /* 104: current segment being written */
934 int64_t dlfs_offset; /* 112: offset in curseg for next partial */
935 int64_t dlfs_lastpseg; /* 120: address of last partial written */
936 uint32_t dlfs_inopf; /* 128: inodes per frag */
937
938/* These are configuration parameters. */
939 uint32_t dlfs_minfree; /* 132: minimum percentage of free blocks */
940
941/* These fields can be computed from the others. */
942 uint64_t dlfs_maxfilesize; /* 136: maximum representable file size */
943 uint32_t dlfs_fsbpseg; /* 144: frags (fsb) per segment */
944 uint32_t dlfs_inopb; /* 148: inodes per block */
945 uint32_t dlfs_ifpb; /* 152: IFILE entries per block */
946 uint32_t dlfs_sepb; /* 156: SEGUSE entries per block */
947 uint32_t dlfs_nindir; /* 160: indirect pointers per block */
948 uint32_t dlfs_nseg; /* 164: number of segments */
949 uint32_t dlfs_nspf; /* 168: number of sectors per fragment */
950 uint32_t dlfs_cleansz; /* 172: cleaner info size in blocks */
951 uint32_t dlfs_segtabsz; /* 176: segment table size in blocks */
952 uint32_t dlfs_bshift; /* 180: calc block number from file offset */
953 uint32_t dlfs_ffshift; /* 184: fast mult/div for frag from file */
954 uint32_t dlfs_fbshift; /* 188: fast mult/div for frag from block */
955 uint64_t dlfs_bmask; /* 192: calc block offset from file offset */
956 uint64_t dlfs_ffmask; /* 200: calc frag offset from file offset */
957 uint64_t dlfs_fbmask; /* 208: calc frag offset from block offset */
958 uint32_t dlfs_blktodb; /* 216: blktodb and dbtoblk shift constant */
959 uint32_t dlfs_sushift; /* 220: fast mult/div for segusage table */
960
961 /* 224: superblock disk offsets */
962 int64_t dlfs_sboffs[LFS_MAXNUMSB];
963
964 int32_t dlfs_maxsymlinklen; /* 304: max len of an internal symlink */
965 uint32_t dlfs_nclean; /* 308: Number of clean segments */
966 u_char dlfs_fsmnt[MNAMELEN]; /* 312: name mounted on */
967 uint16_t dlfs_pflags; /* 402: file system persistent flags */
968 int32_t dlfs_dmeta; /* 404: total number of dirty summaries */
969 uint32_t dlfs_minfreeseg; /* 408: segments not counted in bfree */
970 uint32_t dlfs_sumsize; /* 412: size of summary blocks */
971 uint32_t dlfs_ibsize; /* 416: size of inode blocks */
972 uint32_t dlfs_inodefmt; /* 420: inode format version */
973 uint64_t dlfs_serial; /* 424: serial number */
974 int64_t dlfs_s0addr; /* 432: start of segment 0 */
975 uint64_t dlfs_tstamp; /* 440: time stamp */
976 uint32_t dlfs_interleave; /* 448: segment interleave */
977 uint32_t dlfs_ident; /* 452: per-fs identifier */
978 uint32_t dlfs_fsbtodb; /* 456: fsbtodb and dbtodsb shift constant */
979 uint32_t dlfs_resvseg; /* 460: segments reserved for the cleaner */
980 int8_t dlfs_pad[44]; /* 464: round to 512 bytes */
981/* Checksum -- last valid disk field. */
982 uint32_t dlfs_cksum; /* 508: checksum for superblock checking */
983};
984
985__CTASSERT(__alignof(struct dlfs) == __alignof(struct dlfs64));
986
987/* Type used for the inode bitmap */
988typedef uint32_t lfs_bm_t;
989
990/*
991 * Linked list of segments whose byte count needs updating following a
992 * file truncation.
993 */
994struct segdelta {
995 long segnum;
996 size_t num;
997 LIST_ENTRY(segdelta) list;
998};
999
1000/*
1001 * In-memory super block.
1002 */
1003struct lfs {
1004 union { /* on-disk parameters */
1005 struct dlfs u_32;
1006 struct dlfs64 u_64;
1007 } lfs_dlfs_u;
1008
1009/* These fields are set at mount time and are meaningless on disk. */
1010 unsigned lfs_is64 : 1, /* are we lfs64 or lfs32? */
1011 lfs_dobyteswap : 1, /* are we opposite-endian? */
1012 lfs_hasolddirfmt : 1; /* dir entries have no d_type */
1013
1014 struct segment *lfs_sp; /* current segment being written */
1015 struct vnode *lfs_ivnode; /* vnode for the ifile */
1016 uint32_t lfs_seglock; /* single-thread the segment writer */
1017 pid_t lfs_lockpid; /* pid of lock holder */
1018 lwpid_t lfs_locklwp; /* lwp of lock holder */
1019 uint32_t lfs_iocount; /* number of ios pending */
1020 uint32_t lfs_writer; /* don't allow any dirops to start */
1021 uint32_t lfs_dirops; /* count of active directory ops */
1022 kcondvar_t lfs_diropscv; /* condvar of active directory ops */
1023 uint32_t lfs_dirvcount; /* count of VDIROP nodes in this fs */
1024 uint32_t lfs_doifile; /* Write ifile blocks on next write */
1025 uint32_t lfs_nactive; /* Number of segments since last ckp */
1026 int8_t lfs_fmod; /* super block modified flag */
1027 int8_t lfs_ronly; /* mounted read-only flag */
1028#define LFS_NOTYET 0x01
1029#define LFS_IFDIRTY 0x02
1030#define LFS_WARNED 0x04
1031#define LFS_UNDIROP 0x08
1032 int8_t lfs_flags; /* currently unused flag */
1033 uint16_t lfs_activesb; /* toggle between superblocks */
1034 daddr_t lfs_sbactive; /* disk address of current sb write */
1035 struct vnode *lfs_flushvp; /* vnode being flushed */
1036 int lfs_flushvp_fakevref; /* fake vref count for flushvp */
1037 struct vnode *lfs_unlockvp; /* being inactivated in lfs_segunlock */
1038 uint32_t lfs_diropwait; /* # procs waiting on dirop flush */
1039 size_t lfs_devbsize; /* Device block size */
1040 size_t lfs_devbshift; /* Device block shift */
1041 krwlock_t lfs_fraglock;
1042 krwlock_t lfs_iflock; /* Ifile lock */
1043 kcondvar_t lfs_stopcv; /* Wrap lock */
1044 struct lwp *lfs_stoplwp;
1045 pid_t lfs_rfpid; /* Process ID of roll-forward agent */
1046 int lfs_nadirop; /* number of active dirop nodes */
1047 long lfs_ravail; /* blocks pre-reserved for writing */
1048 long lfs_favail; /* blocks pre-reserved for writing */
1049 struct lfs_res_blk *lfs_resblk; /* Reserved memory for pageout */
1050 TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */
1051 TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */
1052#define LFS_RESHASH_WIDTH 17
1053 LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH];
1054 int lfs_pdflush; /* pagedaemon wants us to flush */
1055 uint32_t **lfs_suflags; /* Segment use flags */
1056#ifdef _KERNEL
1057 struct pool lfs_clpool; /* Pool for struct lfs_cluster */
1058 struct pool lfs_bpppool; /* Pool for bpp */
1059 struct pool lfs_segpool; /* Pool for struct segment */
1060#endif /* _KERNEL */
1061#define LFS_MAX_CLEANIND 64
1062 daddr_t lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */
1063 int lfs_cleanind; /* Index into intervals */
1064 int lfs_sleepers; /* # procs sleeping this fs */
1065 kcondvar_t lfs_sleeperscv;
1066 int lfs_pages; /* dirty pages blaming this fs */
1067 lfs_bm_t *lfs_ino_bitmap; /* Inuse inodes bitmap */
1068 int lfs_nowrap; /* Suspend log wrap */
1069 int lfs_wrappass; /* Allow first log wrap requester to pass */
1070 int lfs_wrapstatus; /* Wrap status */
1071 int lfs_reclino; /* Inode being reclaimed */
1072 daddr_t lfs_startseg; /* Segment we started writing at */
1073 LIST_HEAD(, segdelta) lfs_segdhd; /* List of pending trunc accounting events */
1074
1075#ifdef _KERNEL
1076 /* The block device we're mounted on. */
1077 dev_t lfs_dev;
1078 struct vnode *lfs_devvp;
1079
1080 /* ULFS-level information */
1081 uint32_t um_flags; /* ULFS flags (below) */
1082 u_long um_nindir; /* indirect ptrs per block */
1083 u_long um_lognindir; /* log2 of um_nindir */
1084 u_long um_bptrtodb; /* indir ptr to disk block */
1085 u_long um_seqinc; /* inc between seq blocks */
1086 int um_maxsymlinklen;
1087 int um_dirblksiz;
1088 uint64_t um_maxfilesize;
1089
1090 /* Stuff used by quota2 code, not currently operable */
1091 unsigned lfs_use_quota2 : 1;
1092 uint32_t lfs_quota_magic;
1093 uint8_t lfs_quota_flags;
1094 uint64_t lfs_quotaino[2];
1095
1096 /* Sleep address replacing &lfs_avail inside the on-disk superblock */
1097 /* XXX: should be replaced with a condvar */
1098 int lfs_availsleep;
1099 /* This one replaces &lfs_nextseg... all ditto */
1100 kcondvar_t lfs_nextsegsleep;
1101
1102 /* Cleaner lwp, set on first bmapv syscall. */
1103 struct lwp *lfs_cleaner_thread;
1104
1105 /* Hint from cleaner, only valid if curlwp == um_cleaner_thread. */
1106 /* XXX change this to BLOCK_INFO after resorting this file */
1107 struct block_info *lfs_cleaner_hint;
1108#endif
1109};
1110
1111/*
1112 * Structures used by lfs_bmapv and lfs_markv to communicate information
1113 * about inodes and data blocks.
1114 */
1115typedef struct block_info {
1116 uint64_t bi_inode; /* inode # */
1117 int64_t bi_lbn; /* logical block w/in file */
1118 int64_t bi_daddr; /* disk address of block */
1119 uint64_t bi_segcreate; /* origin segment create time */
1120 int bi_version; /* file version number */
1121 int bi_size; /* size of the block (if fragment) */
1122 void *bi_bp; /* data buffer */
1123} BLOCK_INFO;
1124
1125/* Compatibility for 7.0 binaries */
1126typedef struct block_info_70 {
1127 uint32_t bi_inode; /* inode # */
1128 int32_t bi_lbn; /* logical block w/in file */
1129 int32_t bi_daddr; /* disk address of block */
1130 uint64_t bi_segcreate; /* origin segment create time */
1131 int bi_version; /* file version number */
1132 void *bi_bp; /* data buffer */
1133 int bi_size; /* size of the block (if fragment) */
1134} BLOCK_INFO_70;
1135
1136/* Compatibility for 1.5 binaries */
1137typedef struct block_info_15 {
1138 uint32_t bi_inode; /* inode # */
1139 int32_t bi_lbn; /* logical block w/in file */
1140 int32_t bi_daddr; /* disk address of block */
1141 uint32_t bi_segcreate; /* origin segment create time */
1142 int bi_version; /* file version number */
1143 void *bi_bp; /* data buffer */
1144 int bi_size; /* size of the block (if fragment) */
1145} BLOCK_INFO_15;
1146
1147/*
1148 * 32/64-bit-clean pointer to block pointers. This points into
1149 * already-existing storage; it is mostly used to access the block
1150 * pointers following a FINFO.
1151 */
1152union lfs_blocks {
1153 int64_t *b64;
1154 int32_t *b32;
1155};
1156
1157/* In-memory description of a segment about to be written. */
1158struct segment {
1159 struct lfs *fs; /* file system pointer */
1160 struct buf **bpp; /* pointer to buffer array */
1161 struct buf **cbpp; /* pointer to next available bp */
1162 struct buf **start_bpp; /* pointer to first bp in this set */
1163 struct buf *ibp; /* buffer pointer to inode page */
1164 union lfs_dinode *idp; /* pointer to ifile dinode */
1165 FINFO *fip; /* current fileinfo pointer */
1166 struct vnode *vp; /* vnode being gathered */
1167 void *segsum; /* segment summary info */
1168 uint32_t ninodes; /* number of inodes in this segment */
1169 int32_t seg_bytes_left; /* bytes left in segment */
1170 int32_t sum_bytes_left; /* bytes left in summary block */
1171 uint32_t seg_number; /* number of this segment */
1172 union lfs_blocks start_lbp; /* beginning lbn for this set */
1173
1174#define SEGM_CKP 0x0001 /* doing a checkpoint */
1175#define SEGM_CLEAN 0x0002 /* cleaner call; don't sort */
1176#define SEGM_SYNC 0x0004 /* wait for segment */
1177#define SEGM_PROT 0x0008 /* don't inactivate at segunlock */
1178#define SEGM_PAGEDAEMON 0x0010 /* pagedaemon called us */
1179#define SEGM_WRITERD 0x0020 /* LFS writed called us */
1180#define SEGM_FORCE_CKP 0x0040 /* Force checkpoint right away */
1181#define SEGM_RECLAIM 0x0080 /* Writing to reclaim vnode */
1182#define SEGM_SINGLE 0x0100 /* Opportunistic writevnodes */
1183 uint16_t seg_flags; /* run-time flags for this segment */
1184 uint32_t seg_iocount; /* number of ios pending */
1185 int ndupino; /* number of duplicate inodes */
1186};
1187
1188/* Statistics Counters */
1189struct lfs_stats { /* Must match sysctl list in lfs_vfsops.h ! */
1190 u_int segsused;
1191 u_int psegwrites;
1192 u_int psyncwrites;
1193 u_int pcleanwrites;
1194 u_int blocktot;
1195 u_int cleanblocks;
1196 u_int ncheckpoints;
1197 u_int nwrites;
1198 u_int nsync_writes;
1199 u_int wait_exceeded;
1200 u_int write_exceeded;
1201 u_int flush_invoked;
1202 u_int vflush_invoked;
1203 u_int clean_inlocked;
1204 u_int clean_vnlocked;
1205 u_int segs_reclaimed;
1206};
1207
1208/* Fcntls to take the place of the lfs syscalls */
1209struct lfs_fcntl_markv {
1210 BLOCK_INFO *blkiov; /* blocks to relocate */
1211 int blkcnt; /* number of blocks (limited to 65536) */
1212};
1213
1214#define LFCNSEGWAITALL _FCNR_FSPRIV('L', 14, struct timeval)
1215#define LFCNSEGWAIT _FCNR_FSPRIV('L', 15, struct timeval)
1216#define LFCNBMAPV _FCNRW_FSPRIV('L', 16, struct lfs_fcntl_markv)
1217#define LFCNMARKV _FCNRW_FSPRIV('L', 17, struct lfs_fcntl_markv)
1218#define LFCNRECLAIM _FCNO_FSPRIV('L', 4)
1219
1220struct lfs_fhandle {
1221 char space[28]; /* FHANDLE_SIZE_COMPAT (but used from userland too) */
1222};
1223#define LFCNREWIND _FCNR_FSPRIV('L', 6, int)
1224#define LFCNINVAL _FCNR_FSPRIV('L', 7, int)
1225#define LFCNRESIZE _FCNR_FSPRIV('L', 8, int)
1226#define LFCNWRAPSTOP _FCNR_FSPRIV('L', 9, int)
1227#define LFCNWRAPGO _FCNR_FSPRIV('L', 10, int)
1228#define LFCNIFILEFH _FCNW_FSPRIV('L', 11, struct lfs_fhandle)
1229#define LFCNWRAPPASS _FCNR_FSPRIV('L', 12, int)
1230# define LFS_WRAP_GOING 0x0
1231# define LFS_WRAP_WAITING 0x1
1232#define LFCNWRAPSTATUS _FCNW_FSPRIV('L', 13, int)
1233
1234/* Debug segment lock */
1235#ifdef notyet
1236# define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs))
1237# define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs))
1238# define ASSERT_DUNNO_SEGLOCK(fs)
1239# define ASSERT_MAYBE_SEGLOCK(fs)
1240#else /* !notyet */
1241# define ASSERT_DUNNO_SEGLOCK(fs) \
1242 DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \
1243 LFS_SEGLOCK_HELD(fs)))
1244# define ASSERT_SEGLOCK(fs) do { \
1245 if (!LFS_SEGLOCK_HELD(fs)) { \
1246 DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \
1247 } \
1248} while(0)
1249# define ASSERT_NO_SEGLOCK(fs) do { \
1250 if (LFS_SEGLOCK_HELD(fs)) { \
1251 DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \
1252 } \
1253} while(0)
1254# define ASSERT_MAYBE_SEGLOCK(x)
1255#endif /* !notyet */
1256
1257/*
1258 * Arguments to mount LFS filesystems
1259 */
1260struct ulfs_args {
1261 char *fspec; /* block special device to mount */
1262};
1263
1264__BEGIN_DECLS
1265void lfs_itimes(struct inode *, const struct timespec *,
1266 const struct timespec *, const struct timespec *);
1267__END_DECLS
1268
1269#endif /* !_UFS_LFS_LFS_H_ */