1/* $NetBSD: lfs.h,v 1.199 2016/06/20 03:25:46 dholland 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
359struct lfs_dirheader64 {
360 uint32_t dh_inoA; /* inode number of entry */
361 uint32_t dh_inoB; /* 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};
366
367union lfs_dirheader {
368 struct lfs_dirheader64 u_64;
369 struct lfs_dirheader32 u_32;
370};
371
372typedef union lfs_dirheader LFS_DIRHEADER;
373
374/*
375 * Template for manipulating directories.
376 */
377
378struct lfs_dirtemplate32 {
379 struct lfs_dirheader32 dot_header;
380 char dot_name[4]; /* must be multiple of 4 */
381 struct lfs_dirheader32 dotdot_header;
382 char dotdot_name[4]; /* ditto */
383};
384
385struct lfs_dirtemplate64 {
386 struct lfs_dirheader64 dot_header;
387 char dot_name[4]; /* must be multiple of 4 */
388 struct lfs_dirheader64 dotdot_header;
389 char dotdot_name[4]; /* ditto */
390};
391
392union lfs_dirtemplate {
393 struct lfs_dirtemplate64 u_64;
394 struct lfs_dirtemplate32 u_32;
395};
396
397#if 0
398/*
399 * This is the old format of directories, sans type element.
400 */
401struct lfs_odirtemplate {
402 uint32_t dot_ino;
403 int16_t dot_reclen;
404 uint16_t dot_namlen;
405 char dot_name[4]; /* must be multiple of 4 */
406 uint32_t dotdot_ino;
407 int16_t dotdot_reclen;
408 uint16_t dotdot_namlen;
409 char dotdot_name[4]; /* ditto */
410};
411#endif
412
413/*
414 * Inodes
415 */
416
417/*
418 * A dinode contains all the meta-data associated with a LFS file.
419 * This structure defines the on-disk format of a dinode. Since
420 * this structure describes an on-disk structure, all its fields
421 * are defined by types with precise widths.
422 */
423
424struct lfs32_dinode {
425 uint16_t di_mode; /* 0: IFMT, permissions; see below. */
426 int16_t di_nlink; /* 2: File link count. */
427 uint32_t di_inumber; /* 4: Inode number. */
428 uint64_t di_size; /* 8: File byte count. */
429 int32_t di_atime; /* 16: Last access time. */
430 int32_t di_atimensec; /* 20: Last access time. */
431 int32_t di_mtime; /* 24: Last modified time. */
432 int32_t di_mtimensec; /* 28: Last modified time. */
433 int32_t di_ctime; /* 32: Last inode change time. */
434 int32_t di_ctimensec; /* 36: Last inode change time. */
435 int32_t di_db[ULFS_NDADDR]; /* 40: Direct disk blocks. */
436 int32_t di_ib[ULFS_NIADDR]; /* 88: Indirect disk blocks. */
437 uint32_t di_flags; /* 100: Status flags (chflags). */
438 uint32_t di_blocks; /* 104: Blocks actually held. */
439 int32_t di_gen; /* 108: Generation number. */
440 uint32_t di_uid; /* 112: File owner. */
441 uint32_t di_gid; /* 116: File group. */
442 uint64_t di_modrev; /* 120: i_modrev for NFSv4 */
443};
444
445struct lfs64_dinode {
446 uint16_t di_mode; /* 0: IFMT, permissions; see below. */
447 int16_t di_nlink; /* 2: File link count. */
448 uint32_t di_uid; /* 4: File owner. */
449 uint32_t di_gid; /* 8: File group. */
450 uint32_t di_blksize; /* 12: Inode blocksize. */
451 uint64_t di_size; /* 16: File byte count. */
452 uint64_t di_blocks; /* 24: Bytes actually held. */
453 int64_t di_atime; /* 32: Last access time. */
454 int64_t di_mtime; /* 40: Last modified time. */
455 int64_t di_ctime; /* 48: Last inode change time. */
456 int64_t di_birthtime; /* 56: Inode creation time. */
457 int32_t di_mtimensec; /* 64: Last modified time. */
458 int32_t di_atimensec; /* 68: Last access time. */
459 int32_t di_ctimensec; /* 72: Last inode change time. */
460 int32_t di_birthnsec; /* 76: Inode creation time. */
461 int32_t di_gen; /* 80: Generation number. */
462 uint32_t di_kernflags; /* 84: Kernel flags. */
463 uint32_t di_flags; /* 88: Status flags (chflags). */
464 int32_t di_extsize; /* 92: External attributes block. */
465 int64_t di_extb[ULFS_NXADDR];/* 96: External attributes block. */
466 int64_t di_db[ULFS_NDADDR]; /* 112: Direct disk blocks. */
467 int64_t di_ib[ULFS_NIADDR]; /* 208: Indirect disk blocks. */
468 uint64_t di_modrev; /* 232: i_modrev for NFSv4 */
469 uint64_t di_inumber; /* 240: Inode number */
470 uint64_t di_spare[1]; /* 244: Reserved; currently unused */
471};
472
473union lfs_dinode {
474 struct lfs64_dinode u_64;
475 struct lfs32_dinode u_32;
476};
477
478/*
479 * The di_db fields may be overlaid with other information for
480 * file types that do not have associated disk storage. Block
481 * and character devices overlay the first data block with their
482 * dev_t value. Short symbolic links place their path in the
483 * di_db area.
484 */
485#define di_rdev di_db[0]
486
487/* Size of the on-disk inode. */
488//#define LFS_DINODE1_SIZE (sizeof(struct ulfs1_dinode)) /* 128 */
489//#define LFS_DINODE2_SIZE (sizeof(struct ulfs2_dinode))
490
491/* File types, found in the upper bits of di_mode. */
492#define LFS_IFMT 0170000 /* Mask of file type. */
493#define LFS_IFIFO 0010000 /* Named pipe (fifo). */
494#define LFS_IFCHR 0020000 /* Character device. */
495#define LFS_IFDIR 0040000 /* Directory file. */
496#define LFS_IFBLK 0060000 /* Block device. */
497#define LFS_IFREG 0100000 /* Regular file. */
498#define LFS_IFLNK 0120000 /* Symbolic link. */
499#define LFS_IFSOCK 0140000 /* UNIX domain socket. */
500#define LFS_IFWHT 0160000 /* Whiteout. */
501
502/*
503 * "struct buf" associated definitions
504 */
505
506/* Unassigned disk addresses. */
507#define UNASSIGNED -1
508#define UNWRITTEN -2
509
510/* Unused logical block number */
511#define LFS_UNUSED_LBN -1
512
513/*
514 * "struct inode" associated definitions
515 */
516
517/* For convenience */
518#define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING)
519
520/*
521 * On-disk and in-memory checkpoint segment usage structure.
522 */
523typedef struct segusage SEGUSE;
524struct segusage {
525 uint32_t su_nbytes; /* 0: number of live bytes */
526 uint32_t su_olastmod; /* 4: SEGUSE last modified timestamp */
527 uint16_t su_nsums; /* 8: number of summaries in segment */
528 uint16_t su_ninos; /* 10: number of inode blocks in seg */
529
530#define SEGUSE_ACTIVE 0x01 /* segment currently being written */
531#define SEGUSE_DIRTY 0x02 /* segment has data in it */
532#define SEGUSE_SUPERBLOCK 0x04 /* segment contains a superblock */
533#define SEGUSE_ERROR 0x08 /* cleaner: do not clean segment */
534#define SEGUSE_EMPTY 0x10 /* segment is empty */
535#define SEGUSE_INVAL 0x20 /* segment is invalid */
536 uint32_t su_flags; /* 12: segment flags */
537 uint64_t su_lastmod; /* 16: last modified timestamp */
538};
539
540typedef struct segusage_v1 SEGUSE_V1;
541struct segusage_v1 {
542 uint32_t su_nbytes; /* 0: number of live bytes */
543 uint32_t su_lastmod; /* 4: SEGUSE last modified timestamp */
544 uint16_t su_nsums; /* 8: number of summaries in segment */
545 uint16_t su_ninos; /* 10: number of inode blocks in seg */
546 uint32_t su_flags; /* 12: segment flags */
547};
548
549/*
550 * On-disk file information. One per file with data blocks in the segment.
551 *
552 * The FINFO structure is a header; it is followed by fi_nblocks block
553 * pointers, which are logical block numbers of the file. (These are the
554 * blocks of the file present in this segment.)
555 */
556
557typedef struct finfo64 FINFO64;
558struct finfo64 {
559 uint32_t fi_nblocks; /* number of blocks */
560 uint32_t fi_version; /* version number */
561 uint64_t fi_ino; /* inode number */
562 uint32_t fi_lastlength; /* length of last block in array */
563 uint32_t fi_pad; /* unused */
564};
565
566typedef struct finfo32 FINFO32;
567struct finfo32 {
568 uint32_t fi_nblocks; /* number of blocks */
569 uint32_t fi_version; /* version number */
570 uint32_t fi_ino; /* inode number */
571 uint32_t fi_lastlength; /* length of last block in array */
572};
573
574typedef union finfo {
575 struct finfo64 u_64;
576 struct finfo32 u_32;
577} FINFO;
578
579/*
580 * inode info (part of the segment summary)
581 *
582 * Each one of these is just a block number; wrapping the structure
583 * around it gives more contextual information in the code about
584 * what's going on.
585 */
586
587typedef struct iinfo64 {
588 uint64_t ii_block; /* block number */
589} IINFO64;
590
591typedef struct iinfo32 {
592 uint32_t ii_block; /* block number */
593} IINFO32;
594
595typedef union iinfo {
596 struct iinfo64 u_64;
597 struct iinfo32 u_32;
598} IINFO;
599
600/*
601 * Index file inode entries.
602 */
603
604/* magic value for daddrs */
605#define LFS_UNUSED_DADDR 0 /* out-of-band daddr */
606/* magic value for if_nextfree */
607#define LFS_ORPHAN_NEXTFREE (~(uint32_t)0) /* indicate orphaned file */
608
609typedef struct ifile64 IFILE64;
610struct ifile64 {
611 uint32_t if_version; /* inode version number */
612 uint32_t if_atime_nsec; /* and nanoseconds */
613 uint64_t if_atime_sec; /* Last access time, seconds */
614 int64_t if_daddr; /* inode disk address */
615 uint64_t if_nextfree; /* next-unallocated inode */
616};
617
618typedef struct ifile32 IFILE32;
619struct ifile32 {
620 uint32_t if_version; /* inode version number */
621 int32_t if_daddr; /* inode disk address */
622 uint32_t if_nextfree; /* next-unallocated inode */
623 uint32_t if_atime_sec; /* Last access time, seconds */
624 uint32_t if_atime_nsec; /* and nanoseconds */
625};
626
627typedef struct ifile_v1 IFILE_V1;
628struct ifile_v1 {
629 uint32_t if_version; /* inode version number */
630 int32_t if_daddr; /* inode disk address */
631 uint32_t if_nextfree; /* next-unallocated inode */
632#if LFS_ATIME_IFILE
633#error "this cannot work"
634 struct timespec if_atime; /* Last access time */
635#endif
636};
637
638/*
639 * Note: struct ifile_v1 is often handled by accessing the first three
640 * fields of struct ifile32. (XXX: Blah. This should be cleaned up as
641 * it may in some cases violate the strict-aliasing rules.)
642 */
643typedef union ifile {
644 struct ifile64 u_64;
645 struct ifile32 u_32;
646 struct ifile_v1 u_v1;
647} IFILE;
648
649/*
650 * Cleaner information structure. This resides in the ifile and is used
651 * to pass information from the kernel to the cleaner.
652 */
653
654/* flags for ->flags */
655#define LFS_CLEANER_MUST_CLEAN 0x01
656
657typedef struct _cleanerinfo32 {
658 uint32_t clean; /* 0: number of clean segments */
659 uint32_t dirty; /* 4: number of dirty segments */
660 int32_t bfree; /* 8: disk blocks free */
661 int32_t avail; /* 12: disk blocks available */
662 uint32_t free_head; /* 16: head of the inode free list */
663 uint32_t free_tail; /* 20: tail of the inode free list */
664 uint32_t flags; /* 24: status word from the kernel */
665} CLEANERINFO32;
666
667typedef struct _cleanerinfo64 {
668 uint32_t clean; /* 0: number of clean segments */
669 uint32_t dirty; /* 4: number of dirty segments */
670 int64_t bfree; /* 8: disk blocks free */
671 int64_t avail; /* 16: disk blocks available */
672 uint64_t free_head; /* 24: head of the inode free list */
673 uint64_t free_tail; /* 32: tail of the inode free list */
674 uint32_t flags; /* 40: status word from the kernel */
675 uint32_t pad; /* 44: must be 64-bit aligned */
676} CLEANERINFO64;
677
678/* this must not go to disk directly of course */
679typedef union _cleanerinfo {
680 CLEANERINFO32 u_32;
681 CLEANERINFO64 u_64;
682} CLEANERINFO;
683
684/*
685 * On-disk segment summary information
686 */
687
688/* magic value for ss_magic */
689#define SS_MAGIC 0x061561
690
691/* flags for ss_flags */
692#define SS_DIROP 0x01 /* segment begins a dirop */
693#define SS_CONT 0x02 /* more partials to finish this write*/
694#define SS_CLEAN 0x04 /* written by the cleaner */
695#define SS_RFW 0x08 /* written by the roll-forward agent */
696#define SS_RECLAIM 0x10 /* written by the roll-forward agent */
697
698/* type used for reading checksum signatures from metadata structures */
699typedef uint32_t lfs_checkword;
700
701typedef struct segsum_v1 SEGSUM_V1;
702struct segsum_v1 {
703 uint32_t ss_sumsum; /* 0: check sum of summary block */
704 uint32_t ss_datasum; /* 4: check sum of data */
705 uint32_t ss_magic; /* 8: segment summary magic number */
706 int32_t ss_next; /* 12: next segment */
707 uint32_t ss_create; /* 16: creation time stamp */
708 uint16_t ss_nfinfo; /* 20: number of file info structures */
709 uint16_t ss_ninos; /* 22: number of inodes in summary */
710 uint16_t ss_flags; /* 24: used for directory operations */
711 uint16_t ss_pad; /* 26: extra space */
712 /* FINFO's and inode daddr's... */
713};
714
715typedef struct segsum32 SEGSUM32;
716struct segsum32 {
717 uint32_t ss_sumsum; /* 0: check sum of summary block */
718 uint32_t ss_datasum; /* 4: check sum of data */
719 uint32_t ss_magic; /* 8: segment summary magic number */
720 int32_t ss_next; /* 12: next segment (disk address) */
721 uint32_t ss_ident; /* 16: roll-forward fsid */
722 uint16_t ss_nfinfo; /* 20: number of file info structures */
723 uint16_t ss_ninos; /* 22: number of inodes in summary */
724 uint16_t ss_flags; /* 24: used for directory operations */
725 uint8_t ss_pad[2]; /* 26: extra space */
726 uint32_t ss_reclino; /* 28: inode being reclaimed */
727 uint64_t ss_serial; /* 32: serial number */
728 uint64_t ss_create; /* 40: time stamp */
729 /* FINFO's and inode daddr's... */
730};
731
732typedef struct segsum64 SEGSUM64;
733struct segsum64 {
734 uint32_t ss_sumsum; /* 0: check sum of summary block */
735 uint32_t ss_datasum; /* 4: check sum of data */
736 uint32_t ss_magic; /* 8: segment summary magic number */
737 uint32_t ss_ident; /* 12: roll-forward fsid */
738 int64_t ss_next; /* 16: next segment (disk address) */
739 uint16_t ss_nfinfo; /* 24: number of file info structures */
740 uint16_t ss_ninos; /* 26: number of inodes in summary */
741 uint16_t ss_flags; /* 28: used for directory operations */
742 uint8_t ss_pad[2]; /* 30: extra space */
743 uint64_t ss_reclino; /* 32: inode being reclaimed */
744 uint64_t ss_serial; /* 40: serial number */
745 uint64_t ss_create; /* 48: time stamp */
746 /* FINFO's and inode daddr's... */
747};
748
749typedef union segsum SEGSUM;
750union segsum {
751 struct segsum64 u_64;
752 struct segsum32 u_32;
753 struct segsum_v1 u_v1;
754};
755
756
757/*
758 * On-disk super block.
759 *
760 * We have separate superblock structures for the 32-bit and 64-bit
761 * LFS, and accessor functions to hide the differences.
762 *
763 * For lfs64, the format version is always 2; version 1 lfs is old.
764 * For both, the inode format version is 0; for lfs32 this selects the
765 * same 32-bit inode as always, and for lfs64 this selects the larger
766 * 64-bit inode structure we got from ffsv2.
767 *
768 * In lfs64:
769 * - inode numbers are 64 bit now
770 * - segments may not be larger than 4G (counted in bytes)
771 * - there may not be more than 2^32 (or perhaps 2^31) segments
772 * - the total volume size is limited to 2^63 frags and/or 2^63
773 * disk blocks, and probably in practice 2^63 bytes.
774 */
775
776#define LFS_MAGIC 0x070162
777#define LFS_MAGIC_SWAPPED 0x62010700
778
779#define LFS64_MAGIC (0x19620701 ^ 0xffffffff)
780#define LFS64_MAGIC_SWAPPED (0x01076219 ^ 0xffffffff)
781
782#define LFS_VERSION 2
783
784#define LFS_MIN_SBINTERVAL 5 /* min superblock segment spacing */
785#define LFS_MAXNUMSB 10 /* max number of superblocks */
786
787/* flags for dlfs_pflags */
788#define LFS_PF_CLEAN 0x1
789
790/* Inode format versions */
791#define LFS_44INODEFMT 0
792#define LFS_MAXINODEFMT 0
793
794struct dlfs {
795 uint32_t dlfs_magic; /* 0: magic number */
796 uint32_t dlfs_version; /* 4: version number */
797
798 uint32_t dlfs_size; /* 8: number of blocks in fs (v1) */
799 /* number of frags in fs (v2) */
800 uint32_t dlfs_ssize; /* 12: number of blocks per segment (v1) */
801 /* number of bytes per segment (v2) */
802 uint32_t dlfs_dsize; /* 16: number of disk blocks in fs */
803 uint32_t dlfs_bsize; /* 20: file system block size */
804 uint32_t dlfs_fsize; /* 24: size of frag blocks in fs */
805 uint32_t dlfs_frag; /* 28: number of frags in a block in fs */
806
807/* Checkpoint region. */
808 uint32_t dlfs_freehd; /* 32: start of the free inode list */
809 int32_t dlfs_bfree; /* 36: number of free frags */
810 uint32_t dlfs_nfiles; /* 40: number of allocated inodes */
811 int32_t dlfs_avail; /* 44: blocks available for writing */
812 int32_t dlfs_uinodes; /* 48: inodes in cache not yet on disk */
813 int32_t dlfs_idaddr; /* 52: inode file disk address */
814 uint32_t dlfs_ifile; /* 56: inode file inode number */
815 int32_t dlfs_lastseg; /* 60: address of last segment written */
816 int32_t dlfs_nextseg; /* 64: address of next segment to write */
817 int32_t dlfs_curseg; /* 68: current segment being written */
818 int32_t dlfs_offset; /* 72: offset in curseg for next partial */
819 int32_t dlfs_lastpseg; /* 76: address of last partial written */
820 uint32_t dlfs_inopf; /* 80: v1: time stamp; v2: inodes per frag */
821
822/* These are configuration parameters. */
823 uint32_t dlfs_minfree; /* 84: minimum percentage of free blocks */
824
825/* These fields can be computed from the others. */
826 uint64_t dlfs_maxfilesize; /* 88: maximum representable file size */
827 uint32_t dlfs_fsbpseg; /* 96: frags (fsb) per segment */
828 uint32_t dlfs_inopb; /* 100: inodes per block */
829 uint32_t dlfs_ifpb; /* 104: IFILE entries per block */
830 uint32_t dlfs_sepb; /* 108: SEGUSE entries per block */
831 uint32_t dlfs_nindir; /* 112: indirect pointers per block */
832 uint32_t dlfs_nseg; /* 116: number of segments */
833 uint32_t dlfs_nspf; /* 120: number of sectors per fragment */
834 uint32_t dlfs_cleansz; /* 124: cleaner info size in blocks */
835 uint32_t dlfs_segtabsz; /* 128: segment table size in blocks */
836 uint32_t dlfs_segmask; /* 132: calculate offset within a segment */
837 uint32_t dlfs_segshift; /* 136: fast mult/div for segments */
838 uint32_t dlfs_bshift; /* 140: calc block number from file offset */
839 uint32_t dlfs_ffshift; /* 144: fast mult/div for frag from file */
840 uint32_t dlfs_fbshift; /* 148: fast mult/div for frag from block */
841 uint64_t dlfs_bmask; /* 152: calc block offset from file offset */
842 uint64_t dlfs_ffmask; /* 160: calc frag offset from file offset */
843 uint64_t dlfs_fbmask; /* 168: calc frag offset from block offset */
844 uint32_t dlfs_blktodb; /* 176: blktodb and dbtoblk shift constant */
845 uint32_t dlfs_sushift; /* 180: fast mult/div for segusage table */
846
847 int32_t dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
848 /* 188: superblock disk offsets */
849 int32_t dlfs_sboffs[LFS_MAXNUMSB];
850
851 uint32_t dlfs_nclean; /* 228: Number of clean segments */
852 u_char dlfs_fsmnt[MNAMELEN]; /* 232: name mounted on */
853 uint16_t dlfs_pflags; /* 322: file system persistent flags */
854 int32_t dlfs_dmeta; /* 324: total number of dirty summaries */
855 uint32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */
856 uint32_t dlfs_sumsize; /* 332: size of summary blocks */
857 uint64_t dlfs_serial; /* 336: serial number */
858 uint32_t dlfs_ibsize; /* 344: size of inode blocks */
859 int32_t dlfs_s0addr; /* 348: start of segment 0 */
860 uint64_t dlfs_tstamp; /* 352: time stamp */
861 uint32_t dlfs_inodefmt; /* 360: inode format version */
862 uint32_t dlfs_interleave; /* 364: segment interleave */
863 uint32_t dlfs_ident; /* 368: per-fs identifier */
864 uint32_t dlfs_fsbtodb; /* 372: fsbtodb and dbtodsb shift constant */
865 uint32_t dlfs_resvseg; /* 376: segments reserved for the cleaner */
866 int8_t dlfs_pad[128]; /* 380: round to 512 bytes */
867/* Checksum -- last valid disk field. */
868 uint32_t dlfs_cksum; /* 508: checksum for superblock checking */
869};
870
871struct dlfs64 {
872 uint32_t dlfs_magic; /* 0: magic number */
873 uint32_t dlfs_version; /* 4: version number (2) */
874
875 uint64_t dlfs_size; /* 8: number of frags in fs (v2) */
876 uint64_t dlfs_dsize; /* 16: number of disk blocks in fs */
877 uint32_t dlfs_ssize; /* 24: number of bytes per segment (v2) */
878 uint32_t dlfs_bsize; /* 28: file system block size */
879 uint32_t dlfs_fsize; /* 32: size of frag blocks in fs */
880 uint32_t dlfs_frag; /* 36: number of frags in a block in fs */
881
882/* Checkpoint region. */
883 uint64_t dlfs_freehd; /* 40: start of the free inode list */
884 uint64_t dlfs_nfiles; /* 48: number of allocated inodes */
885 int64_t dlfs_bfree; /* 56: number of free frags */
886 int64_t dlfs_avail; /* 64: blocks available for writing */
887 int64_t dlfs_idaddr; /* 72: inode file disk address */
888 int32_t dlfs_uinodes; /* 80: inodes in cache not yet on disk */
889 uint32_t dlfs_unused_0; /* 84: not used */
890 int64_t dlfs_lastseg; /* 88: address of last segment written */
891 int64_t dlfs_nextseg; /* 96: address of next segment to write */
892 int64_t dlfs_curseg; /* 104: current segment being written */
893 int64_t dlfs_offset; /* 112: offset in curseg for next partial */
894 int64_t dlfs_lastpseg; /* 120: address of last partial written */
895 uint32_t dlfs_inopf; /* 128: inodes per frag */
896
897/* These are configuration parameters. */
898 uint32_t dlfs_minfree; /* 132: minimum percentage of free blocks */
899
900/* These fields can be computed from the others. */
901 uint64_t dlfs_maxfilesize; /* 136: maximum representable file size */
902 uint32_t dlfs_fsbpseg; /* 144: frags (fsb) per segment */
903 uint32_t dlfs_inopb; /* 148: inodes per block */
904 uint32_t dlfs_ifpb; /* 152: IFILE entries per block */
905 uint32_t dlfs_sepb; /* 156: SEGUSE entries per block */
906 uint32_t dlfs_nindir; /* 160: indirect pointers per block */
907 uint32_t dlfs_nseg; /* 164: number of segments */
908 uint32_t dlfs_nspf; /* 168: number of sectors per fragment */
909 uint32_t dlfs_cleansz; /* 172: cleaner info size in blocks */
910 uint32_t dlfs_segtabsz; /* 176: segment table size in blocks */
911 uint32_t dlfs_bshift; /* 180: calc block number from file offset */
912 uint32_t dlfs_ffshift; /* 184: fast mult/div for frag from file */
913 uint32_t dlfs_fbshift; /* 188: fast mult/div for frag from block */
914 uint64_t dlfs_bmask; /* 192: calc block offset from file offset */
915 uint64_t dlfs_ffmask; /* 200: calc frag offset from file offset */
916 uint64_t dlfs_fbmask; /* 208: calc frag offset from block offset */
917 uint32_t dlfs_blktodb; /* 216: blktodb and dbtoblk shift constant */
918 uint32_t dlfs_sushift; /* 220: fast mult/div for segusage table */
919
920 /* 224: superblock disk offsets */
921 int64_t dlfs_sboffs[LFS_MAXNUMSB];
922
923 int32_t dlfs_maxsymlinklen; /* 304: max len of an internal symlink */
924 uint32_t dlfs_nclean; /* 308: Number of clean segments */
925 u_char dlfs_fsmnt[MNAMELEN]; /* 312: name mounted on */
926 uint16_t dlfs_pflags; /* 402: file system persistent flags */
927 int32_t dlfs_dmeta; /* 404: total number of dirty summaries */
928 uint32_t dlfs_minfreeseg; /* 408: segments not counted in bfree */
929 uint32_t dlfs_sumsize; /* 412: size of summary blocks */
930 uint32_t dlfs_ibsize; /* 416: size of inode blocks */
931 uint32_t dlfs_inodefmt; /* 420: inode format version */
932 uint64_t dlfs_serial; /* 424: serial number */
933 int64_t dlfs_s0addr; /* 432: start of segment 0 */
934 uint64_t dlfs_tstamp; /* 440: time stamp */
935 uint32_t dlfs_interleave; /* 448: segment interleave */
936 uint32_t dlfs_ident; /* 452: per-fs identifier */
937 uint32_t dlfs_fsbtodb; /* 456: fsbtodb and dbtodsb shift constant */
938 uint32_t dlfs_resvseg; /* 460: segments reserved for the cleaner */
939 int8_t dlfs_pad[44]; /* 464: round to 512 bytes */
940/* Checksum -- last valid disk field. */
941 uint32_t dlfs_cksum; /* 508: checksum for superblock checking */
942};
943
944/* Type used for the inode bitmap */
945typedef uint32_t lfs_bm_t;
946
947/*
948 * Linked list of segments whose byte count needs updating following a
949 * file truncation.
950 */
951struct segdelta {
952 long segnum;
953 size_t num;
954 LIST_ENTRY(segdelta) list;
955};
956
957/*
958 * In-memory super block.
959 */
960struct lfs {
961 union { /* on-disk parameters */
962 struct dlfs u_32;
963 struct dlfs64 u_64;
964 } lfs_dlfs_u;
965
966/* These fields are set at mount time and are meaningless on disk. */
967 unsigned lfs_is64 : 1, /* are we lfs64 or lfs32? */
968 lfs_dobyteswap : 1, /* are we opposite-endian? */
969 lfs_hasolddirfmt : 1; /* dir entries have no d_type */
970
971 struct segment *lfs_sp; /* current segment being written */
972 struct vnode *lfs_ivnode; /* vnode for the ifile */
973 uint32_t lfs_seglock; /* single-thread the segment writer */
974 pid_t lfs_lockpid; /* pid of lock holder */
975 lwpid_t lfs_locklwp; /* lwp of lock holder */
976 uint32_t lfs_iocount; /* number of ios pending */
977 uint32_t lfs_writer; /* don't allow any dirops to start */
978 uint32_t lfs_dirops; /* count of active directory ops */
979 uint32_t lfs_dirvcount; /* count of VDIROP nodes in this fs */
980 uint32_t lfs_doifile; /* Write ifile blocks on next write */
981 uint32_t lfs_nactive; /* Number of segments since last ckp */
982 int8_t lfs_fmod; /* super block modified flag */
983 int8_t lfs_ronly; /* mounted read-only flag */
984#define LFS_NOTYET 0x01
985#define LFS_IFDIRTY 0x02
986#define LFS_WARNED 0x04
987#define LFS_UNDIROP 0x08
988 int8_t lfs_flags; /* currently unused flag */
989 uint16_t lfs_activesb; /* toggle between superblocks */
990 daddr_t lfs_sbactive; /* disk address of current sb write */
991 struct vnode *lfs_flushvp; /* vnode being flushed */
992 int lfs_flushvp_fakevref; /* fake vref count for flushvp */
993 struct vnode *lfs_unlockvp; /* being inactivated in lfs_segunlock */
994 uint32_t lfs_diropwait; /* # procs waiting on dirop flush */
995 size_t lfs_devbsize; /* Device block size */
996 size_t lfs_devbshift; /* Device block shift */
997 krwlock_t lfs_fraglock;
998 krwlock_t lfs_iflock; /* Ifile lock */
999 kcondvar_t lfs_stopcv; /* Wrap lock */
1000 struct lwp *lfs_stoplwp;
1001 pid_t lfs_rfpid; /* Process ID of roll-forward agent */
1002 int lfs_nadirop; /* number of active dirop nodes */
1003 long lfs_ravail; /* blocks pre-reserved for writing */
1004 long lfs_favail; /* blocks pre-reserved for writing */
1005 struct lfs_res_blk *lfs_resblk; /* Reserved memory for pageout */
1006 TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */
1007 TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */
1008#define LFS_RESHASH_WIDTH 17
1009 LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH];
1010 int lfs_pdflush; /* pagedaemon wants us to flush */
1011 uint32_t **lfs_suflags; /* Segment use flags */
1012#ifdef _KERNEL
1013 struct pool lfs_clpool; /* Pool for struct lfs_cluster */
1014 struct pool lfs_bpppool; /* Pool for bpp */
1015 struct pool lfs_segpool; /* Pool for struct segment */
1016#endif /* _KERNEL */
1017#define LFS_MAX_CLEANIND 64
1018 daddr_t lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */
1019 int lfs_cleanind; /* Index into intervals */
1020 int lfs_sleepers; /* # procs sleeping this fs */
1021 int lfs_pages; /* dirty pages blaming this fs */
1022 lfs_bm_t *lfs_ino_bitmap; /* Inuse inodes bitmap */
1023 int lfs_nowrap; /* Suspend log wrap */
1024 int lfs_wrappass; /* Allow first log wrap requester to pass */
1025 int lfs_wrapstatus; /* Wrap status */
1026 int lfs_reclino; /* Inode being reclaimed */
1027 daddr_t lfs_startseg; /* Segment we started writing at */
1028 LIST_HEAD(, segdelta) lfs_segdhd; /* List of pending trunc accounting events */
1029
1030#ifdef _KERNEL
1031 /* The block device we're mounted on. */
1032 dev_t lfs_dev;
1033 struct vnode *lfs_devvp;
1034
1035 /* ULFS-level information */
1036 uint32_t um_flags; /* ULFS flags (below) */
1037 u_long um_nindir; /* indirect ptrs per block */
1038 u_long um_lognindir; /* log2 of um_nindir */
1039 u_long um_bptrtodb; /* indir ptr to disk block */
1040 u_long um_seqinc; /* inc between seq blocks */
1041 int um_maxsymlinklen;
1042 int um_dirblksiz;
1043 uint64_t um_maxfilesize;
1044
1045 /* Stuff used by quota2 code, not currently operable */
1046 unsigned lfs_use_quota2 : 1;
1047 uint32_t lfs_quota_magic;
1048 uint8_t lfs_quota_flags;
1049 uint64_t lfs_quotaino[2];
1050
1051 /* Sleep address replacing &lfs_avail inside the on-disk superblock */
1052 /* XXX: should be replaced with a condvar */
1053 int lfs_availsleep;
1054 /* This one replaces &lfs_nextseg... all ditto */
1055 int lfs_nextsegsleep;
1056
1057 /* Cleaner lwp, set on first bmapv syscall. */
1058 struct lwp *lfs_cleaner_thread;
1059
1060 /* Hint from cleaner, only valid if curlwp == um_cleaner_thread. */
1061 /* XXX change this to BLOCK_INFO after resorting this file */
1062 struct block_info *lfs_cleaner_hint;
1063#endif
1064};
1065
1066/*
1067 * Structures used by lfs_bmapv and lfs_markv to communicate information
1068 * about inodes and data blocks.
1069 */
1070typedef struct block_info {
1071 uint64_t bi_inode; /* inode # */
1072 int64_t bi_lbn; /* logical block w/in file */
1073 int64_t bi_daddr; /* disk address of block */
1074 uint64_t bi_segcreate; /* origin segment create time */
1075 int bi_version; /* file version number */
1076 int bi_size; /* size of the block (if fragment) */
1077 void *bi_bp; /* data buffer */
1078} BLOCK_INFO;
1079
1080/* Compatibility for 7.0 binaries */
1081typedef struct block_info_70 {
1082 uint32_t bi_inode; /* inode # */
1083 int32_t bi_lbn; /* logical block w/in file */
1084 int32_t bi_daddr; /* disk address of block */
1085 uint64_t bi_segcreate; /* origin segment create time */
1086 int bi_version; /* file version number */
1087 void *bi_bp; /* data buffer */
1088 int bi_size; /* size of the block (if fragment) */
1089} BLOCK_INFO_70;
1090
1091/* Compatibility for 1.5 binaries */
1092typedef struct block_info_15 {
1093 uint32_t bi_inode; /* inode # */
1094 int32_t bi_lbn; /* logical block w/in file */
1095 int32_t bi_daddr; /* disk address of block */
1096 uint32_t bi_segcreate; /* origin segment create time */
1097 int bi_version; /* file version number */
1098 void *bi_bp; /* data buffer */
1099 int bi_size; /* size of the block (if fragment) */
1100} BLOCK_INFO_15;
1101
1102/*
1103 * 32/64-bit-clean pointer to block pointers. This points into
1104 * already-existing storage; it is mostly used to access the block
1105 * pointers following a FINFO.
1106 */
1107union lfs_blocks {
1108 int64_t *b64;
1109 int32_t *b32;
1110};
1111
1112/* In-memory description of a segment about to be written. */
1113struct segment {
1114 struct lfs *fs; /* file system pointer */
1115 struct buf **bpp; /* pointer to buffer array */
1116 struct buf **cbpp; /* pointer to next available bp */
1117 struct buf **start_bpp; /* pointer to first bp in this set */
1118 struct buf *ibp; /* buffer pointer to inode page */
1119 union lfs_dinode *idp; /* pointer to ifile dinode */
1120 FINFO *fip; /* current fileinfo pointer */
1121 struct vnode *vp; /* vnode being gathered */
1122 void *segsum; /* segment summary info */
1123 uint32_t ninodes; /* number of inodes in this segment */
1124 int32_t seg_bytes_left; /* bytes left in segment */
1125 int32_t sum_bytes_left; /* bytes left in summary block */
1126 uint32_t seg_number; /* number of this segment */
1127 union lfs_blocks start_lbp; /* beginning lbn for this set */
1128
1129#define SEGM_CKP 0x0001 /* doing a checkpoint */
1130#define SEGM_CLEAN 0x0002 /* cleaner call; don't sort */
1131#define SEGM_SYNC 0x0004 /* wait for segment */
1132#define SEGM_PROT 0x0008 /* don't inactivate at segunlock */
1133#define SEGM_PAGEDAEMON 0x0010 /* pagedaemon called us */
1134#define SEGM_WRITERD 0x0020 /* LFS writed called us */
1135#define SEGM_FORCE_CKP 0x0040 /* Force checkpoint right away */
1136#define SEGM_RECLAIM 0x0080 /* Writing to reclaim vnode */
1137#define SEGM_SINGLE 0x0100 /* Opportunistic writevnodes */
1138 uint16_t seg_flags; /* run-time flags for this segment */
1139 uint32_t seg_iocount; /* number of ios pending */
1140 int ndupino; /* number of duplicate inodes */
1141};
1142
1143/* Statistics Counters */
1144struct lfs_stats { /* Must match sysctl list in lfs_vfsops.h ! */
1145 u_int segsused;
1146 u_int psegwrites;
1147 u_int psyncwrites;
1148 u_int pcleanwrites;
1149 u_int blocktot;
1150 u_int cleanblocks;
1151 u_int ncheckpoints;
1152 u_int nwrites;
1153 u_int nsync_writes;
1154 u_int wait_exceeded;
1155 u_int write_exceeded;
1156 u_int flush_invoked;
1157 u_int vflush_invoked;
1158 u_int clean_inlocked;
1159 u_int clean_vnlocked;
1160 u_int segs_reclaimed;
1161};
1162
1163/* Fcntls to take the place of the lfs syscalls */
1164struct lfs_fcntl_markv {
1165 BLOCK_INFO *blkiov; /* blocks to relocate */
1166 int blkcnt; /* number of blocks (limited to 65536) */
1167};
1168
1169#define LFCNSEGWAITALL _FCNR_FSPRIV('L', 14, struct timeval)
1170#define LFCNSEGWAIT _FCNR_FSPRIV('L', 15, struct timeval)
1171#define LFCNBMAPV _FCNRW_FSPRIV('L', 16, struct lfs_fcntl_markv)
1172#define LFCNMARKV _FCNRW_FSPRIV('L', 17, struct lfs_fcntl_markv)
1173#define LFCNRECLAIM _FCNO_FSPRIV('L', 4)
1174
1175struct lfs_fhandle {
1176 char space[28]; /* FHANDLE_SIZE_COMPAT (but used from userland too) */
1177};
1178#define LFCNREWIND _FCNR_FSPRIV('L', 6, int)
1179#define LFCNINVAL _FCNR_FSPRIV('L', 7, int)
1180#define LFCNRESIZE _FCNR_FSPRIV('L', 8, int)
1181#define LFCNWRAPSTOP _FCNR_FSPRIV('L', 9, int)
1182#define LFCNWRAPGO _FCNR_FSPRIV('L', 10, int)
1183#define LFCNIFILEFH _FCNW_FSPRIV('L', 11, struct lfs_fhandle)
1184#define LFCNWRAPPASS _FCNR_FSPRIV('L', 12, int)
1185# define LFS_WRAP_GOING 0x0
1186# define LFS_WRAP_WAITING 0x1
1187#define LFCNWRAPSTATUS _FCNW_FSPRIV('L', 13, int)
1188
1189/* Debug segment lock */
1190#ifdef notyet
1191# define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs))
1192# define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs))
1193# define ASSERT_DUNNO_SEGLOCK(fs)
1194# define ASSERT_MAYBE_SEGLOCK(fs)
1195#else /* !notyet */
1196# define ASSERT_DUNNO_SEGLOCK(fs) \
1197 DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \
1198 LFS_SEGLOCK_HELD(fs)))
1199# define ASSERT_SEGLOCK(fs) do { \
1200 if (!LFS_SEGLOCK_HELD(fs)) { \
1201 DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \
1202 } \
1203} while(0)
1204# define ASSERT_NO_SEGLOCK(fs) do { \
1205 if (LFS_SEGLOCK_HELD(fs)) { \
1206 DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \
1207 } \
1208} while(0)
1209# define ASSERT_MAYBE_SEGLOCK(x)
1210#endif /* !notyet */
1211
1212/*
1213 * Arguments to mount LFS filesystems
1214 */
1215struct ulfs_args {
1216 char *fspec; /* block special device to mount */
1217};
1218
1219__BEGIN_DECLS
1220void lfs_itimes(struct inode *, const struct timespec *,
1221 const struct timespec *, const struct timespec *);
1222__END_DECLS
1223
1224#endif /* !_UFS_LFS_LFS_H_ */
1225