Home | History | Annotate | Line # | Download | only in lfs
      1 /*	$NetBSD: lfs.h,v 1.213 2025/11/06 15:45:32 perseant 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 (at) 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 
    214 /*
    215  * Directories
    216  */
    217 
    218 /*
    219  * Directories in LFS are files; they use the same inode and block
    220  * mapping structures that regular files do. The directory per se is
    221  * manifested in the file contents: an unordered, unstructured
    222  * sequence of variable-size directory entries.
    223  *
    224  * This format and structure is taken (via what was originally shared
    225  * ufs-level code) from FFS. Each directory entry is a fixed header
    226  * followed by a string, the total length padded to a 4-byte boundary.
    227  * All strings include a null terminator; the maximum string length
    228  * is LFS_MAXNAMLEN, which is 255.
    229  *
    230  * The directory entry header structure (struct lfs_dirheader) is just
    231  * the header information. A complete entry is this plus a null-
    232  * terminated name following it, plus some amount of padding. The
    233  * length of the name (not including the null terminator) is given by
    234  * the namlen field of the header; the complete record length,
    235  * including the null terminator and padding, is given by the reclen
    236  * field of the header. The record length is always 4-byte aligned.
    237  * (Even on 64-bit volumes, the record length is only 4-byte aligned,
    238  * not 8-byte.)
    239  *
    240  * Historically, FFS directories were/are organized into blocks of
    241  * size DIRBLKSIZE that can be written atomically to disk at the
    242  * hardware level. Directory entries are not allowed to cross the
    243  * boundaries of these blocks. The resulting atomicity is important
    244  * for the integrity of FFS volumes; however, for LFS it's irrelevant.
    245  * All we have to care about is not writing out directories that
    246  * confuse earlier ufs-based versions of the LFS code.
    247  *
    248  * This means [to be determined]. (XXX)
    249  *
    250  * As DIRBLKSIZE in its FFS sense is hardware-dependent, and file
    251  * system images do from time to time move to different hardware, code
    252  * that reads directories should be prepared to handle directories
    253  * written in a context where DIRBLKSIZE was different (smaller or
    254  * larger) than its current value. Note however that it is not
    255  * sensible for DIRBLKSIZE to be larger than the volume fragment size,
    256  * and not practically possible for it to be larger than the volume
    257  * block size.
    258  *
    259  * Some further notes:
    260  *    - the LFS_DIRSIZ macro provides the minimum space needed to hold
    261  *      a directory entry.
    262  *    - any particular entry may be arbitrarily larger (which is why the
    263  *      header stores both the entry size and the name size) to pad out
    264  *      unused space.
    265  *    - historically the padding in an entry is not necessarily zeroed
    266  *      but may contain trash.
    267  *    - dp->d_reclen is the size of the entry. This is always 4-byte
    268  *      aligned.
    269  *    - dp->d_namlen is the length of the string, and should always be
    270  *      the same as strlen(dp->d_name).
    271  *    - in particular, space available in an entry is given by
    272  *      dp->d_reclen - LFS_DIRSIZ(dp), and all space available within a
    273  *      directory block is tucked away within an existing entry.
    274  *    - all space within a directory block is part of some entry.
    275  *    - therefore, inserting a new entry requires finding and
    276  *      splitting a suitable existing entry, and when entries are
    277  *      removed their space is merged into the entry ahead of them.
    278  *    - an empty/unused entry has d_ino set to 0. This normally only
    279  *      appears in the first entry in a block, as elsewhere the unused
    280  *      entry should have been merged into the one before it. However,
    281  *      fsck leaves such entries behind so they must be tolerated
    282  *      elsewhere.
    283  *    - a completely empty directory block has one entry whose
    284  *      d_reclen is DIRBLKSIZ and whose d_ino is 0.
    285  *
    286  * The "old directory format" referenced by the fs->lfs_isolddirfmt
    287  * flag (and some other things) refers to when the type field was
    288  * added to directory entries. This change was made to FFS in the 80s,
    289  * well before LFS was first written; there should be no LFS volumes
    290  * (and certainly no LFS v2-format volumes or LFS64 volumes) where the
    291  * old format pertains. All of the related logic should probably be
    292  * removed; however, it hasn't been yet, and we get to carry it around
    293  * until we can be conclusively sure it isn't needed.
    294  *
    295  * In the "old directory format" there is no type field and the namlen
    296  * field is correspondingly 16 bits wide. On big-endian volumes this
    297  * has no effect: namlen cannot exceed 255, so the upper byte is
    298  * always 0 and this reads back from the type field as LFS_DT_UNKNOWN.
    299  * On little-endian volumes, the namlen field will always be 0 and
    300  * the namlen value needs to be read out of the type field. (The type
    301  * is always LFS_DT_UNKNOWN.) The directory accessor functions take
    302  * care of this so nothing else needs to be aware of it.
    303  *
    304  * LFS_OLDDIRFMT and LFS_NEWDIRFMT are code numbers for the old and
    305  * new directory format respectively. These codes do not appear on
    306  * disk; they're generated from a runtime macro called FSFMT() that's
    307  * cued by other things. This is why (confusingly) LFS_OLDDIRFMT is 1
    308  * and LFS_NEWDIRFMT is 0.
    309  *
    310  * FSFMT(), LFS_OLDDIRFMT, and LFS_NEWDIRFMT should be removed. (XXX)
    311  */
    312 
    313 /*
    314  * Directory block size.
    315  */
    316 #undef	LFS_DIRBLKSIZ
    317 #define	LFS_DIRBLKSIZ	DEV_BSIZE
    318 
    319 /*
    320  * Convert between stat structure type codes and directory entry type codes.
    321  */
    322 #define	LFS_IFTODT(mode)	(((mode) & 0170000) >> 12)
    323 #define	LFS_DTTOIF(dirtype)	((dirtype) << 12)
    324 
    325 /*
    326  * Theoretically, directories can be more than 2Gb in length; however, in
    327  * practice this seems unlikely. So, we define the type doff_t as a 32-bit
    328  * quantity to keep down the cost of doing lookup on a 32-bit machine.
    329  */
    330 #define	doff_t		int32_t
    331 #define	lfs_doff_t	int32_t
    332 #define	LFS_MAXDIRSIZE	(0x7fffffff)
    333 
    334 /*
    335  * File types for d_type
    336  */
    337 #define	LFS_DT_UNKNOWN	 0
    338 #define	LFS_DT_FIFO	 1
    339 #define	LFS_DT_CHR	 2
    340 #define	LFS_DT_DIR	 4
    341 #define	LFS_DT_BLK	 6
    342 #define	LFS_DT_REG	 8
    343 #define	LFS_DT_LNK	10
    344 #define	LFS_DT_SOCK	12
    345 #define	LFS_DT_WHT	14
    346 
    347 /*
    348  * (See notes above)
    349  */
    350 
    351 struct lfs_dirheader32 {
    352 	uint32_t dh_ino;		/* inode number of entry */
    353 	uint16_t dh_reclen;		/* length of this record */
    354 	uint8_t  dh_type; 		/* file type, see below */
    355 	uint8_t  dh_namlen;		/* length of string in d_name */
    356 };
    357 __CTASSERT(sizeof(struct lfs_dirheader32) == 8);
    358 
    359 struct lfs_dirheader64 {
    360 	uint64_t dh_ino;		/* inode number of entry */
    361 	uint16_t dh_reclen;		/* length of this record */
    362 	uint8_t  dh_type; 		/* file type, see below */
    363 	uint8_t  dh_namlen;		/* length of string in d_name */
    364 } __aligned(4) __packed;
    365 __CTASSERT(sizeof(struct lfs_dirheader64) == 12);
    366 
    367 union lfs_dirheader {
    368 	struct lfs_dirheader64 u_64;
    369 	struct lfs_dirheader32 u_32;
    370 };
    371 __CTASSERT(__alignof(union lfs_dirheader) == __alignof(struct lfs_dirheader64));
    372 #ifndef __lint__
    373 __CTASSERT(__alignof(union lfs_dirheader) == __alignof(struct lfs_dirheader32));
    374 #endif
    375 
    376 typedef union lfs_dirheader LFS_DIRHEADER;
    377 
    378 /*
    379  * Template for manipulating directories.
    380  */
    381 
    382 struct lfs_dirtemplate32 {
    383 	struct lfs_dirheader32	dot_header;
    384 	char			dot_name[4];	/* must be multiple of 4 */
    385 	struct lfs_dirheader32	dotdot_header;
    386 	char			dotdot_name[4];	/* ditto */
    387 };
    388 __CTASSERT(sizeof(struct lfs_dirtemplate32) == 2*(8 + 4));
    389 
    390 struct lfs_dirtemplate64 {
    391 	struct lfs_dirheader64	dot_header;
    392 	char			dot_name[4];	/* must be multiple of 4 */
    393 	struct lfs_dirheader64	dotdot_header;
    394 	char			dotdot_name[4];	/* ditto */
    395 };
    396 __CTASSERT(sizeof(struct lfs_dirtemplate64) == 2*(12 + 4));
    397 
    398 union lfs_dirtemplate {
    399 	struct lfs_dirtemplate64 u_64;
    400 	struct lfs_dirtemplate32 u_32;
    401 };
    402 
    403 #if 0
    404 /*
    405  * This is the old format of directories, sans type element.
    406  */
    407 struct lfs_odirtemplate {
    408 	uint32_t	dot_ino;
    409 	int16_t		dot_reclen;
    410 	uint16_t	dot_namlen;
    411 	char		dot_name[4];	/* must be multiple of 4 */
    412 	uint32_t	dotdot_ino;
    413 	int16_t		dotdot_reclen;
    414 	uint16_t	dotdot_namlen;
    415 	char		dotdot_name[4];	/* ditto */
    416 };
    417 __CTASSERT(sizeof(struct lfs_odirtemplate) == 2*(8 + 4));
    418 #endif
    419 
    420 /*
    421  * Inodes
    422  */
    423 
    424 /*
    425  * A dinode contains all the meta-data associated with a LFS file.
    426  * This structure defines the on-disk format of a dinode. Since
    427  * this structure describes an on-disk structure, all its fields
    428  * are defined by types with precise widths.
    429  */
    430 
    431 struct lfs32_dinode {
    432 	uint16_t	di_mode;	/*   0: IFMT, permissions; see below. */
    433 	int16_t		di_nlink;	/*   2: File link count. */
    434 	uint32_t	di_inumber;	/*   4: Inode number. */
    435 	uint64_t	di_size;	/*   8: File byte count. */
    436 	int32_t		di_atime;	/*  16: Last access time. */
    437 	int32_t		di_atimensec;	/*  20: Last access time. */
    438 	int32_t		di_mtime;	/*  24: Last modified time. */
    439 	int32_t		di_mtimensec;	/*  28: Last modified time. */
    440 	int32_t		di_ctime;	/*  32: Last inode change time. */
    441 	int32_t		di_ctimensec;	/*  36: Last inode change time. */
    442 	int32_t		di_db[ULFS_NDADDR]; /*  40: Direct disk blocks. */
    443 	int32_t		di_ib[ULFS_NIADDR]; /*  88: Indirect disk blocks. */
    444 	uint32_t	di_flags;	/* 100: Status flags (chflags). */
    445 	uint32_t	di_blocks;	/* 104: Blocks actually held. */
    446 	int32_t		di_gen;		/* 108: Generation number. */
    447 	uint32_t	di_uid;		/* 112: File owner. */
    448 	uint32_t	di_gid;		/* 116: File group. */
    449 	uint64_t	di_modrev;	/* 120: i_modrev for NFSv4 */
    450 };
    451 __CTASSERT(sizeof(struct lfs32_dinode) == 128);
    452 
    453 struct lfs64_dinode {
    454 	uint16_t	di_mode;	/*   0: IFMT, permissions; see below. */
    455 	int16_t		di_nlink;	/*   2: File link count. */
    456 	uint32_t	di_uid;		/*   4: File owner. */
    457 	uint32_t	di_gid;		/*   8: File group. */
    458 	uint32_t	di_blksize;	/*  12: Inode blocksize. */
    459 	uint64_t	di_size;	/*  16: File byte count. */
    460 	uint64_t	di_blocks;	/*  24: Bytes actually held. */
    461 	int64_t		di_atime;	/*  32: Last access time. */
    462 	int64_t		di_mtime;	/*  40: Last modified time. */
    463 	int64_t		di_ctime;	/*  48: Last inode change time. */
    464 	int64_t		di_birthtime;	/*  56: Inode creation time. */
    465 	int32_t		di_mtimensec;	/*  64: Last modified time. */
    466 	int32_t		di_atimensec;	/*  68: Last access time. */
    467 	int32_t		di_ctimensec;	/*  72: Last inode change time. */
    468 	int32_t		di_birthnsec;	/*  76: Inode creation time. */
    469 	int32_t		di_gen;		/*  80: Generation number. */
    470 	uint32_t	di_kernflags;	/*  84: Kernel flags. */
    471 	uint32_t	di_flags;	/*  88: Status flags (chflags). */
    472 	int32_t		di_extsize;	/*  92: External attributes block. */
    473 	int64_t		di_extb[ULFS_NXADDR];/* 96: External attributes block. */
    474 	int64_t		di_db[ULFS_NDADDR]; /* 112: Direct disk blocks. */
    475 	int64_t		di_ib[ULFS_NIADDR]; /* 208: Indirect disk blocks. */
    476 	uint64_t	di_modrev;	/* 232: i_modrev for NFSv4 */
    477 	uint64_t	di_inumber;	/* 240: Inode number */
    478 	uint64_t	di_spare[1];	/* 248: Reserved; currently unused */
    479 };
    480 __CTASSERT(sizeof(struct lfs64_dinode) == 256);
    481 
    482 union lfs_dinode {
    483 	struct lfs64_dinode u_64;
    484 	struct lfs32_dinode u_32;
    485 };
    486 __CTASSERT(__alignof(union lfs_dinode) == __alignof(struct lfs64_dinode));
    487 __CTASSERT(__alignof(union lfs_dinode) == __alignof(struct lfs32_dinode));
    488 
    489 /*
    490  * The di_db fields may be overlaid with other information for
    491  * file types that do not have associated disk storage. Block
    492  * and character devices overlay the first data block with their
    493  * dev_t value. Short symbolic links place their path in the
    494  * di_db area.
    495  */
    496 #define	di_rdev		di_db[0]
    497 
    498 /* Size of the on-disk inode. */
    499 //#define	LFS_DINODE1_SIZE	(sizeof(struct ulfs1_dinode))	/* 128 */
    500 //#define	LFS_DINODE2_SIZE	(sizeof(struct ulfs2_dinode))
    501 
    502 /* File types, found in the upper bits of di_mode. */
    503 #define	LFS_IFMT	0170000		/* Mask of file type. */
    504 #define	LFS_IFIFO	0010000		/* Named pipe (fifo). */
    505 #define	LFS_IFCHR	0020000		/* Character device. */
    506 #define	LFS_IFDIR	0040000		/* Directory file. */
    507 #define	LFS_IFBLK	0060000		/* Block device. */
    508 #define	LFS_IFREG	0100000		/* Regular file. */
    509 #define	LFS_IFLNK	0120000		/* Symbolic link. */
    510 #define	LFS_IFSOCK	0140000		/* UNIX domain socket. */
    511 #define	LFS_IFWHT	0160000		/* Whiteout. */
    512 
    513 /*
    514  * "struct buf" associated definitions
    515  */
    516 
    517 /* Unassigned disk addresses. */
    518 #define	UNASSIGNED	-1
    519 #define UNWRITTEN	-2
    520 
    521 /* Unused logical block number */
    522 #define LFS_UNUSED_LBN	-1
    523 
    524 /*
    525  * On-disk and in-memory checkpoint segment usage structure.
    526  */
    527 typedef struct segusage SEGUSE;
    528 struct segusage {
    529 	uint32_t su_nbytes;		/* 0: number of live bytes */
    530 	uint32_t su_olastmod;		/* 4: SEGUSE last modified timestamp */
    531 	uint16_t su_nsums;		/* 8: number of summaries in segment */
    532 	uint16_t su_ninos;		/* 10: number of inode blocks in seg */
    533 
    534 #define	SEGUSE_ACTIVE		0x01	/*  segment currently being written */
    535 #define	SEGUSE_DIRTY		0x02	/*  segment has data in it */
    536 #define	SEGUSE_SUPERBLOCK	0x04	/*  segment contains a superblock */
    537 #define SEGUSE_ERROR		0x08	/*  cleaner: do not clean segment */
    538 #define SEGUSE_EMPTY		0x10	/*  segment is empty */
    539 #define SEGUSE_INVAL		0x20	/*  segment is invalid */
    540 #define SEGUSE_READY		0x40	/*  segment is ready for reclaim */
    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 
    546 typedef struct segusage_v1 SEGUSE_V1;
    547 struct 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 
    564 typedef struct finfo64 FINFO64;
    565 struct 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 
    574 typedef struct finfo32 FINFO32;
    575 struct 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 
    583 typedef 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 
    600 typedef struct iinfo64 {
    601 	uint64_t ii_block;		/* block number */
    602 } __aligned(4) __packed IINFO64;
    603 __CTASSERT(sizeof(struct iinfo64) == 8);
    604 
    605 typedef struct iinfo32 {
    606 	uint32_t ii_block;		/* block number */
    607 } IINFO32;
    608 __CTASSERT(sizeof(struct iinfo32) == 4);
    609 
    610 typedef 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 values for daddrs */
    624 #define	LFS_UNUSED_DADDR	0	/* out-of-band daddr */
    625 #ifdef DEBUG
    626 # define LFS_ILLEGAL_DADDR	-1	/* out-of-band daddr */
    627 # define DADDR_IS_BAD(daddr) \
    628 	((daddr) == LFS_UNUSED_DADDR || (daddr) == LFS_ILLEGAL_DADDR)
    629 #else /* ! DEBUG */
    630 # define LFS_ILLEGAL_DADDR	LFS_UNUSED_DADDR
    631 # define DADDR_IS_BAD(daddr) ((daddr) == LFS_UNUSED_DADDR)
    632 #endif /* ! DEBUG */
    633 
    634 /* magic value for if_nextfree -- indicate orphaned file */
    635 #define LFS_ORPHAN_NEXTFREE(fs) \
    636 	((fs)->lfs_is64 ? ~(uint64_t)0 : ~(uint32_t)0)
    637 
    638 #define INUM_IS_BAD(fs, ino) \
    639 	(ino == LFS_UNUSED_INUM || ino == LFS_ORPHAN_NEXTFREE(fs))
    640 
    641 typedef struct ifile64 IFILE64;
    642 struct ifile64 {
    643 	uint32_t if_version;		/* inode version number */
    644 	uint32_t if_atime_nsec;		/* and nanoseconds */
    645 	uint64_t if_atime_sec;		/* Last access time, seconds */
    646 	int64_t	  if_daddr;		/* inode disk address */
    647 	uint64_t if_nextfree;		/* next-unallocated inode */
    648 } __aligned(4) __packed;
    649 __CTASSERT(sizeof(struct ifile64) == 32);
    650 
    651 typedef struct ifile32 IFILE32;
    652 struct ifile32 {
    653 	uint32_t if_version;		/* inode version number */
    654 	int32_t	  if_daddr;		/* inode disk address */
    655 	uint32_t if_nextfree;		/* next-unallocated inode */
    656 	uint32_t if_atime_sec;		/* Last access time, seconds */
    657 	uint32_t if_atime_nsec;		/* and nanoseconds */
    658 };
    659 __CTASSERT(sizeof(struct ifile32) == 20);
    660 
    661 typedef struct ifile_v1 IFILE_V1;
    662 struct ifile_v1 {
    663 	uint32_t if_version;		/* inode version number */
    664 	int32_t	  if_daddr;		/* inode disk address */
    665 	uint32_t if_nextfree;		/* next-unallocated inode */
    666 #if LFS_ATIME_IFILE
    667 #error "this cannot work"
    668 	struct timespec if_atime;	/* Last access time */
    669 #endif
    670 };
    671 __CTASSERT(sizeof(struct ifile_v1) == 12);
    672 
    673 /*
    674  * Note: struct ifile_v1 is often handled by accessing the first three
    675  * fields of struct ifile32. (XXX: Blah.  This should be cleaned up as
    676  * it may in some cases violate the strict-aliasing rules.)
    677  */
    678 typedef union ifile {
    679 	struct ifile64 u_64;
    680 	struct ifile32 u_32;
    681 	struct ifile_v1 u_v1;
    682 } IFILE;
    683 __CTASSERT(__alignof(union ifile) == __alignof(struct ifile64));
    684 #ifndef __lint__
    685 __CTASSERT(__alignof(union ifile) == __alignof(struct ifile32));
    686 __CTASSERT(__alignof(union ifile) == __alignof(struct ifile_v1));
    687 #endif
    688 
    689 /*
    690  * Cleaner information structure.  This resides in the ifile and is used
    691  * to pass information from the kernel to the cleaner.
    692  */
    693 
    694 /* flags for ->flags */
    695 #define LFS_CLEANER_MUST_CLEAN	0x01
    696 
    697 typedef struct _cleanerinfo32 {
    698 	uint32_t clean;			/* 0: number of clean segments */
    699 	uint32_t dirty;			/* 4: number of dirty segments */
    700 	int32_t   bfree;		/* 8: disk blocks free */
    701 	int32_t	  avail;		/* 12: disk blocks available */
    702 	uint32_t free_head;		/* 16: head of the inode free list */
    703 	uint32_t free_tail;		/* 20: tail of the inode free list */
    704 	uint32_t flags;			/* 24: status word from the kernel */
    705 } CLEANERINFO32;
    706 __CTASSERT(sizeof(struct _cleanerinfo32) == 28);
    707 
    708 typedef struct _cleanerinfo64 {
    709 	uint32_t clean;			/* 0: number of clean segments */
    710 	uint32_t dirty;			/* 4: number of dirty segments */
    711 	int64_t   bfree;		/* 8: disk blocks free */
    712 	int64_t	  avail;		/* 16: disk blocks available */
    713 	uint64_t free_head;		/* 24: head of the inode free list */
    714 	uint64_t free_tail;		/* 32: tail of the inode free list */
    715 	uint32_t flags;			/* 40: status word from the kernel */
    716 	uint32_t pad;			/* 44: must be 64-bit aligned */
    717 } __aligned(4) __packed CLEANERINFO64;
    718 __CTASSERT(sizeof(struct _cleanerinfo64) == 48);
    719 
    720 /* this must not go to disk directly of course */
    721 typedef union _cleanerinfo {
    722 	CLEANERINFO32 u_32;
    723 	CLEANERINFO64 u_64;
    724 } CLEANERINFO;
    725 #ifndef __lint__
    726 __CTASSERT(__alignof(union _cleanerinfo) == __alignof(struct _cleanerinfo32));
    727 __CTASSERT(__alignof(union _cleanerinfo) == __alignof(struct _cleanerinfo64));
    728 #endif
    729 
    730 /*
    731  * On-disk segment summary information
    732  */
    733 
    734 /* magic value for ss_magic */
    735 #define SS_MAGIC	0x061561
    736 
    737 /* flags for ss_flags */
    738 #define	SS_DIROP	0x01		/* segment begins a dirop */
    739 #define	SS_CONT		0x02		/* more partials to finish this write*/
    740 #define	SS_CLEAN	0x04		/* written by the cleaner */
    741 #define	SS_RFW		0x08		/* written by the roll-forward agent */
    742 #define	SS_RECLAIM	0x10		/* written by the roll-forward agent */
    743 
    744 /* type used for reading checksum signatures from metadata structures */
    745 typedef uint32_t lfs_checkword;
    746 
    747 typedef struct segsum_v1 SEGSUM_V1;
    748 struct segsum_v1 {
    749 	uint32_t ss_sumsum;		/* 0: check sum of summary block */
    750 	uint32_t ss_datasum;		/* 4: check sum of data */
    751 	uint32_t ss_magic;		/* 8: segment summary magic number */
    752 	int32_t	  ss_next;		/* 12: next segment */
    753 	uint32_t ss_create;		/* 16: creation time stamp */
    754 	uint16_t ss_nfinfo;		/* 20: number of file info structures */
    755 	uint16_t ss_ninos;		/* 22: number of inodes in summary */
    756 	uint16_t ss_flags;		/* 24: used for directory operations */
    757 	uint16_t ss_pad;		/* 26: extra space */
    758 	/* FINFO's and inode daddr's... */
    759 };
    760 __CTASSERT(sizeof(struct segsum_v1) == 28);
    761 
    762 typedef struct segsum32 SEGSUM32;
    763 struct segsum32 {
    764 	uint32_t ss_sumsum;		/* 0: check sum of summary block */
    765 	uint32_t ss_datasum;		/* 4: check sum of data */
    766 	uint32_t ss_magic;		/* 8: segment summary magic number */
    767 	int32_t	  ss_next;		/* 12: next segment (disk address) */
    768 	uint32_t ss_ident;		/* 16: roll-forward fsid */
    769 	uint16_t ss_nfinfo;		/* 20: number of file info structures */
    770 	uint16_t ss_ninos;		/* 22: number of inodes in summary */
    771 	uint16_t ss_flags;		/* 24: used for directory operations */
    772 	uint8_t  ss_pad[2];		/* 26: extra space */
    773 	uint32_t ss_reclino;		/* 28: inode being reclaimed */
    774 	uint64_t ss_serial;		/* 32: serial number */
    775 	uint64_t ss_create;		/* 40: time stamp */
    776 	/* FINFO's and inode daddr's... */
    777 } __aligned(4) __packed;
    778 __CTASSERT(sizeof(struct segsum32) == 48);
    779 
    780 typedef struct segsum64 SEGSUM64;
    781 struct segsum64 {
    782 	uint32_t ss_sumsum;		/* 0: check sum of summary block */
    783 	uint32_t ss_datasum;		/* 4: check sum of data */
    784 	uint32_t ss_magic;		/* 8: segment summary magic number */
    785 	uint32_t ss_ident;		/* 12: roll-forward fsid */
    786 	int64_t	  ss_next;		/* 16: next segment (disk address) */
    787 	uint16_t ss_nfinfo;		/* 24: number of file info structures */
    788 	uint16_t ss_ninos;		/* 26: number of inodes in summary */
    789 	uint16_t ss_flags;		/* 28: used for directory operations */
    790 	uint8_t  ss_pad[2];		/* 30: extra space */
    791 	uint64_t ss_reclino;		/* 32: inode being reclaimed */
    792 	uint64_t ss_serial;		/* 40: serial number */
    793 	uint64_t ss_create;		/* 48: time stamp */
    794 	/* FINFO's and inode daddr's... */
    795 } __aligned(4) __packed;
    796 __CTASSERT(sizeof(struct segsum64) == 56);
    797 
    798 typedef union segsum SEGSUM;
    799 union segsum {
    800 	struct segsum64 u_64;
    801 	struct segsum32 u_32;
    802 	struct segsum_v1 u_v1;
    803 };
    804 __CTASSERT(__alignof(union segsum) == __alignof(struct segsum64));
    805 __CTASSERT(__alignof(union segsum) == __alignof(struct segsum32));
    806 #ifndef __lint__
    807 __CTASSERT(__alignof(union segsum) == __alignof(struct segsum_v1));
    808 #endif
    809 
    810 /*
    811  * On-disk super block.
    812  *
    813  * We have separate superblock structures for the 32-bit and 64-bit
    814  * LFS, and accessor functions to hide the differences.
    815  *
    816  * For lfs64, the format version is always 2; version 1 lfs is old.
    817  * For both, the inode format version is 0; for lfs32 this selects the
    818  * same 32-bit inode as always, and for lfs64 this selects the larger
    819  * 64-bit inode structure we got from ffsv2.
    820  *
    821  * In lfs64:
    822  *   - inode numbers are 64 bit now
    823  *   - segments may not be larger than 4G (counted in bytes)
    824  *   - there may not be more than 2^32 (or perhaps 2^31) segments
    825  *   - the total volume size is limited to 2^63 frags and/or 2^63
    826  *     disk blocks, and probably in practice 2^63 bytes.
    827  */
    828 
    829 #define	       LFS_MAGIC       		0x070162
    830 #define        LFS_MAGIC_SWAPPED	0x62010700
    831 
    832 #define        LFS64_MAGIC     		(0x19620701 ^ 0xffffffff)
    833 #define        LFS64_MAGIC_SWAPPED      (0x01076219 ^ 0xffffffff)
    834 
    835 #define	       LFS_VERSION     		2
    836 
    837 #define LFS_MIN_SBINTERVAL     5	/* min superblock segment spacing */
    838 #define LFS_MAXNUMSB	       10	/* max number of superblocks */
    839 
    840 /* flags for dlfs_pflags */
    841 #define LFS_PF_CLEAN 0x1
    842 
    843 /* Inode format versions */
    844 #define LFS_44INODEFMT 0
    845 #define LFS_MAXINODEFMT 0
    846 
    847 struct dlfs {
    848 	uint32_t dlfs_magic;	  /* 0: magic number */
    849 	uint32_t dlfs_version;	  /* 4: version number */
    850 
    851 	uint32_t dlfs_size;	  /* 8: number of blocks in fs (v1) */
    852 				  /*	number of frags in fs (v2) */
    853 	uint32_t dlfs_ssize;	  /* 12: number of blocks per segment (v1) */
    854 				  /*	 number of bytes per segment (v2) */
    855 	uint32_t dlfs_dsize;	  /* 16: number of disk blocks in fs */
    856 	uint32_t dlfs_bsize;	  /* 20: file system block size */
    857 	uint32_t dlfs_fsize;	  /* 24: size of frag blocks in fs */
    858 	uint32_t dlfs_frag;	  /* 28: number of frags in a block in fs */
    859 
    860 /* Checkpoint region. */
    861 	uint32_t dlfs_freehd;	  /* 32: start of the free inode list */
    862 	int32_t   dlfs_bfree;	  /* 36: number of free frags */
    863 	uint32_t dlfs_nfiles;	  /* 40: number of allocated inodes */
    864 	int32_t	  dlfs_avail;	  /* 44: blocks available for writing */
    865 	int32_t	  dlfs_uinodes;	  /* 48: inodes in cache not yet on disk */
    866 	int32_t	  dlfs_idaddr;	  /* 52: inode file disk address */
    867 	uint32_t dlfs_ifile;	  /* 56: inode file inode number */
    868 	int32_t	  dlfs_lastseg;	  /* 60: address of last segment written */
    869 	int32_t	  dlfs_nextseg;	  /* 64: address of next segment to write */
    870 	int32_t	  dlfs_curseg;	  /* 68: current segment being written */
    871 	int32_t	  dlfs_offset;	  /* 72: offset in curseg for next partial */
    872 	int32_t	  dlfs_lastpseg;  /* 76: address of last partial written */
    873 	uint32_t dlfs_inopf;	  /* 80: v1: time stamp; v2: inodes per frag */
    874 
    875 /* These are configuration parameters. */
    876 	uint32_t dlfs_minfree;	  /* 84: minimum percentage of free blocks */
    877 
    878 /* These fields can be computed from the others. */
    879 	uint64_t dlfs_maxfilesize; /* 88: maximum representable file size */
    880 	uint32_t dlfs_fsbpseg;	  /* 96: frags (fsb) per segment */
    881 	uint32_t dlfs_inopb;	  /* 100: inodes per block */
    882 	uint32_t dlfs_ifpb;	  /* 104: IFILE entries per block */
    883 	uint32_t dlfs_sepb;	  /* 108: SEGUSE entries per block */
    884 	uint32_t dlfs_nindir;	  /* 112: indirect pointers per block */
    885 	uint32_t dlfs_nseg;	  /* 116: number of segments */
    886 	uint32_t dlfs_nspf;	  /* 120: number of sectors per fragment */
    887 	uint32_t dlfs_cleansz;	  /* 124: cleaner info size in blocks */
    888 	uint32_t dlfs_segtabsz;	  /* 128: segment table size in blocks */
    889 	uint32_t dlfs_segmask;	  /* 132: calculate offset within a segment */
    890 	uint32_t dlfs_segshift;	  /* 136: fast mult/div for segments */
    891 	uint32_t dlfs_bshift;	  /* 140: calc block number from file offset */
    892 	uint32_t dlfs_ffshift;	  /* 144: fast mult/div for frag from file */
    893 	uint32_t dlfs_fbshift;	  /* 148: fast mult/div for frag from block */
    894 	uint64_t dlfs_bmask;	  /* 152: calc block offset from file offset */
    895 	uint64_t dlfs_ffmask;	  /* 160: calc frag offset from file offset */
    896 	uint64_t dlfs_fbmask;	  /* 168: calc frag offset from block offset */
    897 	uint32_t dlfs_blktodb;	  /* 176: blktodb and dbtoblk shift constant */
    898 	uint32_t dlfs_sushift;	  /* 180: fast mult/div for segusage table */
    899 
    900 	int32_t	  dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
    901 				  /* 188: superblock disk offsets */
    902 	int32_t	  dlfs_sboffs[LFS_MAXNUMSB];
    903 
    904 	uint32_t dlfs_nclean;	  /* 228: Number of clean segments */
    905 	u_char	  dlfs_fsmnt[MNAMELEN];	 /* 232: name mounted on */
    906 	uint16_t dlfs_pflags;	  /* 322: file system persistent flags */
    907 	int32_t	  dlfs_dmeta;	  /* 324: total number of dirty summaries */
    908 	uint32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */
    909 	uint32_t dlfs_sumsize;	  /* 332: size of summary blocks */
    910 	uint64_t dlfs_serial;	  /* 336: serial number */
    911 	uint32_t dlfs_ibsize;	  /* 344: size of inode blocks */
    912 	int32_t	  dlfs_s0addr;	  /* 348: start of segment 0 */
    913 	uint64_t dlfs_tstamp;	  /* 352: time stamp */
    914 	uint32_t dlfs_inodefmt;	  /* 360: inode format version */
    915 	uint32_t dlfs_interleave; /* 364: segment interleave */
    916 	uint32_t dlfs_ident;	  /* 368: per-fs identifier */
    917 	uint32_t dlfs_fsbtodb;	  /* 372: fsbtodb and dbtodsb shift constant */
    918 	uint32_t dlfs_resvseg;	  /* 376: segments reserved for the cleaner */
    919 	int8_t	  dlfs_pad[128];  /* 380: round to 512 bytes */
    920 /* Checksum -- last valid disk field. */
    921 	uint32_t dlfs_cksum;	  /* 508: checksum for superblock checking */
    922 };
    923 
    924 struct dlfs64 {
    925 	uint32_t dlfs_magic;	  /* 0: magic number */
    926 	uint32_t dlfs_version;	  /* 4: version number (2) */
    927 
    928 	uint64_t dlfs_size;	  /* 8: number of frags in fs (v2) */
    929 	uint64_t dlfs_dsize;	  /* 16: number of disk blocks in fs */
    930 	uint32_t dlfs_ssize;	  /* 24: number of bytes per segment (v2) */
    931 	uint32_t dlfs_bsize;	  /* 28: file system block size */
    932 	uint32_t dlfs_fsize;	  /* 32: size of frag blocks in fs */
    933 	uint32_t dlfs_frag;	  /* 36: number of frags in a block in fs */
    934 
    935 /* Checkpoint region. */
    936 	uint64_t dlfs_freehd;	  /* 40: start of the free inode list */
    937 	uint64_t dlfs_nfiles;	  /* 48: number of allocated inodes */
    938 	int64_t   dlfs_bfree;	  /* 56: number of free frags */
    939 	int64_t	  dlfs_avail;	  /* 64: blocks available for writing */
    940 	int64_t	  dlfs_idaddr;	  /* 72: inode file disk address */
    941 	int32_t	  dlfs_uinodes;	  /* 80: inodes in cache not yet on disk */
    942 	uint32_t dlfs_unused_0;	  /* 84: not used */
    943 	int64_t	  dlfs_lastseg;	  /* 88: address of last segment written */
    944 	int64_t	  dlfs_nextseg;	  /* 96: address of next segment to write */
    945 	int64_t	  dlfs_curseg;	  /* 104: current segment being written */
    946 	int64_t	  dlfs_offset;	  /* 112: offset in curseg for next partial */
    947 	int64_t	  dlfs_lastpseg;  /* 120: address of last partial written */
    948 	uint32_t dlfs_inopf;	  /* 128: inodes per frag */
    949 
    950 /* These are configuration parameters. */
    951 	uint32_t dlfs_minfree;	  /* 132: minimum percentage of free blocks */
    952 
    953 /* These fields can be computed from the others. */
    954 	uint64_t dlfs_maxfilesize; /* 136: maximum representable file size */
    955 	uint32_t dlfs_fsbpseg;	  /* 144: frags (fsb) per segment */
    956 	uint32_t dlfs_inopb;	  /* 148: inodes per block */
    957 	uint32_t dlfs_ifpb;	  /* 152: IFILE entries per block */
    958 	uint32_t dlfs_sepb;	  /* 156: SEGUSE entries per block */
    959 	uint32_t dlfs_nindir;	  /* 160: indirect pointers per block */
    960 	uint32_t dlfs_nseg;	  /* 164: number of segments */
    961 	uint32_t dlfs_nspf;	  /* 168: number of sectors per fragment */
    962 	uint32_t dlfs_cleansz;	  /* 172: cleaner info size in blocks */
    963 	uint32_t dlfs_segtabsz;	  /* 176: segment table size in blocks */
    964 	uint32_t dlfs_bshift;	  /* 180: calc block number from file offset */
    965 	uint32_t dlfs_ffshift;	  /* 184: fast mult/div for frag from file */
    966 	uint32_t dlfs_fbshift;	  /* 188: fast mult/div for frag from block */
    967 	uint64_t dlfs_bmask;	  /* 192: calc block offset from file offset */
    968 	uint64_t dlfs_ffmask;	  /* 200: calc frag offset from file offset */
    969 	uint64_t dlfs_fbmask;	  /* 208: calc frag offset from block offset */
    970 	uint32_t dlfs_blktodb;	  /* 216: blktodb and dbtoblk shift constant */
    971 	uint32_t dlfs_sushift;	  /* 220: fast mult/div for segusage table */
    972 
    973 				  /* 224: superblock disk offsets */
    974 	int64_t	   dlfs_sboffs[LFS_MAXNUMSB];
    975 
    976 	int32_t	  dlfs_maxsymlinklen; /* 304: max len of an internal symlink */
    977 	uint32_t dlfs_nclean;	  /* 308: Number of clean segments */
    978 	u_char	  dlfs_fsmnt[MNAMELEN];	 /* 312: name mounted on */
    979 	uint16_t dlfs_pflags;	  /* 402: file system persistent flags */
    980 	int32_t	  dlfs_dmeta;	  /* 404: total number of dirty summaries */
    981 	uint32_t dlfs_minfreeseg; /* 408: segments not counted in bfree */
    982 	uint32_t dlfs_sumsize;	  /* 412: size of summary blocks */
    983 	uint32_t dlfs_ibsize;	  /* 416: size of inode blocks */
    984 	uint32_t dlfs_inodefmt;	  /* 420: inode format version */
    985 	uint64_t dlfs_serial;	  /* 424: serial number */
    986 	int64_t	  dlfs_s0addr;	  /* 432: start of segment 0 */
    987 	uint64_t dlfs_tstamp;	  /* 440: time stamp */
    988 	uint32_t dlfs_interleave; /* 448: segment interleave */
    989 	uint32_t dlfs_ident;	  /* 452: per-fs identifier */
    990 	uint32_t dlfs_fsbtodb;	  /* 456: fsbtodb and dbtodsb shift constant */
    991 	uint32_t dlfs_resvseg;	  /* 460: segments reserved for the cleaner */
    992 	int8_t	  dlfs_pad[44];   /* 464: round to 512 bytes */
    993 /* Checksum -- last valid disk field. */
    994 	uint32_t dlfs_cksum;	  /* 508: checksum for superblock checking */
    995 };
    996 
    997 __CTASSERT(__alignof(struct dlfs) == __alignof(struct dlfs64));
    998 
    999 /* Type used for the inode bitmap */
   1000 typedef uint32_t lfs_bm_t;
   1001 
   1002 /*
   1003  * Linked list of segments whose byte count needs updating following a
   1004  * file truncation.
   1005  */
   1006 struct segdelta {
   1007 	long segnum;
   1008 	size_t num;
   1009 	LIST_ENTRY(segdelta) list;
   1010 };
   1011 
   1012 /*
   1013  * Parameters used by the in-kernel cleaner.
   1014  */
   1015 struct lfs_autoclean_params {
   1016 	size_t size;  /* Size of this structure */
   1017 #define LFS_CLEANMODE_NONE	0x00 /* Auto-cleaner off */
   1018 #define LFS_CLEANMODE_GREEDY	0x01 /* Greedy selector */
   1019 #define LFS_CLEANMODE_CB	0x02 /* Cost-benefit selector */
   1020 	int mode;     /* Which of the autoclean modes to use */
   1021 	int thresh;   /* Low cleaning threshold to use */
   1022 	int target;   /* How many blocks to rewrite before checkpointing */
   1023 } __packed;
   1024 
   1025 /*
   1026  * In-memory super block.
   1027  */
   1028 struct lfs {
   1029 	union {				/* on-disk parameters */
   1030 		struct dlfs u_32;
   1031 		struct dlfs64 u_64;
   1032 	} lfs_dlfs_u;
   1033 
   1034 /* These fields are set at mount time and are meaningless on disk. */
   1035 	unsigned lfs_is64 : 1,		/* are we lfs64 or lfs32? */
   1036 		lfs_dobyteswap : 1,	/* are we opposite-endian? */
   1037 		lfs_hasolddirfmt : 1;	/* dir entries have no d_type */
   1038 
   1039 	struct segment *lfs_sp;		/* current segment being written */
   1040 	struct vnode *lfs_ivnode;	/* vnode for the ifile */
   1041 	uint32_t  lfs_seglock;		/* single-thread the segment writer */
   1042 	pid_t	  lfs_lockpid;		/* pid of lock holder */
   1043 	lwpid_t	  lfs_locklwp;		/* lwp of lock holder */
   1044 	uint32_t lfs_iocount;		/* number of ios pending */
   1045 	uint32_t lfs_writer;		/* don't allow any dirops to start */
   1046 	uint32_t lfs_dirops;		/* count of active directory ops */
   1047 	kcondvar_t lfs_diropscv;	/* condvar of active directory ops */
   1048 	uint32_t lfs_dirvcount;		/* count of VDIROP nodes in this fs */
   1049 	uint32_t lfs_doifile;		/* Write ifile blocks on next write */
   1050 	uint32_t lfs_nactive;		/* Number of segments since last ckp */
   1051 	int8_t	  lfs_fmod;		/* super block modified flag */
   1052 	int8_t	  lfs_ronly;		/* mounted read-only flag */
   1053 #define LFS_NOTYET	0x01
   1054 #define LFS_IFDIRTY	0x02
   1055 #define LFS_WARNED	0x04
   1056 #define LFS_UNDIROP	0x08
   1057 #define LFS_MUSTCLEAN	0x10
   1058 	int8_t	  lfs_flags;		/* non-persistent flags */
   1059 	uint16_t lfs_activesb;		/* toggle between superblocks */
   1060 	daddr_t	  lfs_sbactive;		/* disk address of current sb write */
   1061 	struct vnode *lfs_flushvp;	/* vnode being flushed */
   1062 	int lfs_flushvp_fakevref;	/* fake vref count for flushvp */
   1063 	struct vnode *lfs_unlockvp;	/* being inactivated in lfs_segunlock */
   1064 	uint32_t lfs_diropwait;		/* # procs waiting on dirop flush */
   1065 	size_t lfs_devbsize;		/* Device block size */
   1066 	size_t lfs_devbshift;		/* Device block shift */
   1067 	krwlock_t lfs_fraglock;
   1068 	krwlock_t lfs_iflock;		/* Ifile lock */
   1069 	kcondvar_t lfs_stopcv;		/* Wrap lock */
   1070 	struct lwp *lfs_stoplwp;
   1071 	pid_t lfs_rfpid;		/* Process ID of roll-forward agent */
   1072 	int	  lfs_nadirop;		/* number of active dirop nodes */
   1073 	long	  lfs_ravail;		/* blocks pre-reserved for writing */
   1074 	long	  lfs_favail;		/* blocks pre-reserved for writing */
   1075 	struct lfs_res_blk *lfs_resblk;	/* Reserved memory for pageout */
   1076 	TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */
   1077 	TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */
   1078 	TAILQ_HEAD(, inode) lfs_cleanhd; /* vnodes being cleaned */
   1079 #define LFS_RESHASH_WIDTH 17
   1080 	LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH];
   1081 	int	  lfs_pdflush;		/* pagedaemon wants us to flush */
   1082 #ifdef _KERNEL
   1083 	struct pool lfs_clpool;		/* Pool for struct lfs_cluster */
   1084 	struct pool lfs_bpppool;	/* Pool for bpp */
   1085 	struct pool lfs_segpool;	/* Pool for struct segment */
   1086 #endif /* _KERNEL */
   1087 #define LFS_MAX_CLEANIND 64
   1088 	daddr_t  lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */
   1089 	int 	 lfs_cleanind;		/* Index into intervals */
   1090 	int lfs_sleepers;		/* # procs sleeping this fs */
   1091 	kcondvar_t lfs_sleeperscv;
   1092 	int lfs_pages;			/* dirty pages blaming this fs */
   1093 	lfs_bm_t *lfs_ino_bitmap;	/* Inuse inodes bitmap */
   1094 	int lfs_nowrap;			/* Suspend log wrap */
   1095 	int lfs_wrappass;		/* Allow first log wrap requester to pass */
   1096 	int lfs_wrapstatus;		/* Wrap status */
   1097 	int lfs_reclino;		/* Inode being reclaimed */
   1098 	daddr_t lfs_startseg;           /* Segment we started writing at */
   1099 	LIST_HEAD(, segdelta) lfs_segdhd;	/* List of pending trunc accounting events */
   1100 
   1101 #ifdef _KERNEL
   1102 	/* The block device we're mounted on. */
   1103 	dev_t lfs_dev;
   1104 	struct vnode *lfs_devvp;
   1105 
   1106 	/* ULFS-level information */
   1107 	uint32_t um_flags;			/* ULFS flags (below) */
   1108 	u_long	um_nindir;			/* indirect ptrs per block */
   1109 	u_long	um_lognindir;			/* log2 of um_nindir */
   1110 	u_long	um_bptrtodb;			/* indir ptr to disk block */
   1111 	u_long	um_seqinc;			/* inc between seq blocks */
   1112 	int um_maxsymlinklen;
   1113 	int um_dirblksiz;
   1114 	uint64_t um_maxfilesize;
   1115 
   1116 	/* Stuff used by quota2 code, not currently operable */
   1117 	unsigned lfs_use_quota2 : 1;
   1118 	uint32_t lfs_quota_magic;
   1119 	uint8_t lfs_quota_flags;
   1120 	uint64_t lfs_quotaino[2];
   1121 
   1122 	/* Sleep address replacing &lfs_avail inside the on-disk superblock */
   1123 	/* XXX: should be replaced with a condvar */
   1124 	int lfs_availsleep;
   1125 	/* This one replaces &lfs_nextseg... all ditto */
   1126 	kcondvar_t lfs_nextsegsleep;
   1127 
   1128 	/* Cleaner lwp, set on first bmapv syscall. */
   1129 	struct lwp *lfs_cleaner_thread;
   1130 
   1131 	/* Hint from cleaner, only valid if curlwp == um_cleaner_thread. */
   1132 	/* XXX change this to BLOCK_INFO after resorting this file */
   1133 	struct block_info *lfs_cleaner_hint;
   1134 
   1135 	/* Cleaner lock */
   1136 	struct lwp *lfs_cleanlock;	/* LWP of lock holder */
   1137 	kcondvar_t lfs_cleanercv;	/* condvar for cleaner */
   1138 
   1139 	/* Cleaner selection function and params */
   1140 	struct lfs_autoclean_params lfs_autoclean;
   1141 	long (*lfs_clean_selector)(struct lfs *, int, SEGUSE *);
   1142 	uint64_t lfs_clean_accum;
   1143 
   1144 	/* Last cleaned segment */
   1145 	uint32_t lfs_lastcleaned;
   1146 #endif
   1147 };
   1148 
   1149 /*
   1150  * Structures used to pass information between the partial-segment
   1151  * parser and its callback functions.
   1152  */
   1153 struct lfs_inofuncarg {
   1154 	struct lfs *fs;		/* The filesystem */
   1155 	daddr_t offset;		/* Address of the inode block */
   1156 	kauth_cred_t cred;	/* Credential */
   1157 	struct lwp *l;		/* lwp */
   1158 	char *buf;		/* Pre-allocated buffer space */
   1159 	void *arg;		/* Opaque argument to function */
   1160 };
   1161 
   1162 struct lfs_finfofuncarg {
   1163 	struct lfs *fs;		/* The file system */
   1164 	FINFO *finfop;		/* The FINFO structure to parse */
   1165 	daddr_t *offsetp;	/* Address of the first block */
   1166 	kauth_cred_t cred;	/* Credential */
   1167 	struct lwp *l;		/* lwp */
   1168 	void *arg;		/* Opaque argument to function */
   1169 };
   1170 
   1171 /*
   1172  * Structures used by lfs_bmapv and lfs_markv to communicate information
   1173  * about inodes and data blocks.
   1174  */
   1175 typedef struct block_info {
   1176 	uint64_t bi_inode;		/* inode # */
   1177 	int64_t	bi_lbn;			/* logical block w/in file */
   1178 	int64_t	bi_daddr;		/* disk address of block */
   1179 	uint64_t bi_segcreate;		/* origin segment create time */
   1180 	int	bi_version;		/* file version number */
   1181 	int	bi_size;		/* size of the block (if fragment) */
   1182 	void	*bi_bp;			/* data buffer */
   1183 } BLOCK_INFO;
   1184 
   1185 /* Compatibility for 7.0 binaries */
   1186 typedef struct block_info_70 {
   1187 	uint32_t bi_inode;		/* inode # */
   1188 	int32_t	bi_lbn;			/* logical block w/in file */
   1189 	int32_t	bi_daddr;		/* disk address of block */
   1190 	uint64_t bi_segcreate;		/* origin segment create time */
   1191 	int	bi_version;		/* file version number */
   1192 	void	*bi_bp;			/* data buffer */
   1193 	int	bi_size;		/* size of the block (if fragment) */
   1194 } BLOCK_INFO_70;
   1195 
   1196 /* Compatibility for 1.5 binaries */
   1197 typedef struct block_info_15 {
   1198 	uint32_t bi_inode;		/* inode # */
   1199 	int32_t	bi_lbn;			/* logical block w/in file */
   1200 	int32_t	bi_daddr;		/* disk address of block */
   1201 	uint32_t bi_segcreate;		/* origin segment create time */
   1202 	int	bi_version;		/* file version number */
   1203 	void	*bi_bp;			/* data buffer */
   1204 	int	bi_size;		/* size of the block (if fragment) */
   1205 } BLOCK_INFO_15;
   1206 
   1207 /*
   1208  * 32/64-bit-clean pointer to block pointers. This points into
   1209  * already-existing storage; it is mostly used to access the block
   1210  * pointers following a FINFO.
   1211  */
   1212 union lfs_blocks {
   1213 	int64_t *b64;
   1214 	int32_t *b32;
   1215 };
   1216 
   1217 /* In-memory description of a segment about to be written. */
   1218 struct segment {
   1219 	struct lfs	 *fs;		/* file system pointer */
   1220 	struct buf	**bpp;		/* pointer to buffer array */
   1221 	struct buf	**cbpp;		/* pointer to next available bp */
   1222 	struct buf	**start_bpp;	/* pointer to first bp in this set */
   1223 	struct buf	 *ibp;		/* buffer pointer to inode page */
   1224 	union lfs_dinode *idp;          /* pointer to ifile dinode */
   1225 	FINFO *fip;			/* current fileinfo pointer */
   1226 	struct vnode	 *vp;		/* vnode being gathered */
   1227 	void	 *segsum;		/* segment summary info */
   1228 	uint32_t ninodes;		/* number of inodes in this segment */
   1229 	int32_t seg_bytes_left;		/* bytes left in segment */
   1230 	int32_t sum_bytes_left;		/* bytes left in summary block */
   1231 	uint32_t seg_number;		/* number of this segment */
   1232 	union lfs_blocks start_lbp;	/* beginning lbn for this set */
   1233 
   1234 #define SEGM_CKP	0x0001		/* doing a checkpoint */
   1235 #define SEGM_CLEAN	0x0002		/* cleaner call; don't sort */
   1236 #define SEGM_SYNC	0x0004		/* wait for segment */
   1237 #define SEGM_PROT	0x0008		/* don't inactivate at segunlock */
   1238 #define SEGM_PAGEDAEMON	0x0010		/* pagedaemon called us */
   1239 #define SEGM_WRITERD	0x0020		/* LFS writed called us */
   1240 #define SEGM_FORCE_CKP	0x0040		/* Force checkpoint right away */
   1241 #define SEGM_RECLAIM	0x0080		/* Writing to reclaim vnode */
   1242 #define SEGM_SINGLE	0x0100		/* Opportunistic writevnodes */
   1243 	uint16_t seg_flags;		/* run-time flags for this segment */
   1244 	uint32_t seg_iocount;		/* number of ios pending */
   1245 	int	  ndupino;		/* number of duplicate inodes */
   1246 	uint64_t bytes_written;		/* bytes written */
   1247 	int	gatherblock_loopcount;	/* lfs_gatherblock loop count */
   1248 };
   1249 
   1250 /* Statistics Counters */
   1251 struct lfs_stats {	/* Must match sysctl list in lfs_vfsops.h ! */
   1252 	u_int	segsused;
   1253 	u_int	psegwrites;
   1254 	u_int	psyncwrites;
   1255 	u_int	pcleanwrites;
   1256 	u_int	blocktot;
   1257 	u_int	cleanblocks;
   1258 	u_int	ncheckpoints;
   1259 	u_int	nwrites;
   1260 	u_int	nsync_writes;
   1261 	u_int	wait_exceeded;
   1262 	u_int	write_exceeded;
   1263 	u_int	flush_invoked;
   1264 	u_int	vflush_invoked;
   1265 	u_int	clean_inlocked;
   1266 	u_int	clean_vnlocked;
   1267 	u_int   segs_reclaimed;
   1268 };
   1269 
   1270 
   1271 /*
   1272  * fcntl-related definitions.
   1273  */
   1274 
   1275 /*
   1276  * Struct lfs_write_stats is returned by many of the LFS fcntl calls.
   1277  * It contains the amount of data written directly by the call, as
   1278  * well as the total amount ths current write offset advanced, both
   1279  * measured in "fsb" (fragments).
   1280  */
   1281 struct lfs_write_stats {
   1282 	int direct; /* Direct data written, in frags */
   1283 	int offset; /* Offset difference, in frags */
   1284 };
   1285 
   1286 /*
   1287  * Update the current segment offset somewhere before the given
   1288  * segment number.  Primarily used for the old, manual file system
   1289  * resize method.
   1290  */
   1291 #define LFCNREWIND	_FCNR_FSPRIV('L', 6, int)
   1292 
   1293 /*
   1294  * Mark the given segment as temporarily invalid, that is, it should
   1295  * not be written to unless no other clean segments are available.
   1296  * Primarily used for the old, manual file system resize method.
   1297  */
   1298 #define LFCNINVAL	_FCNR_FSPRIV('L', 7, int)
   1299 
   1300 /*
   1301  * Resize the file system to the given number of segments.  If the
   1302  * file system is being shrunk, enough clean segments must already
   1303  * exist to accommodate the size change.
   1304  */
   1305 #define LFCNRESIZE	_FCNR_FSPRIV('L', 8, int)
   1306 
   1307 /*
   1308  * Wrap control.
   1309  *
   1310  * LFCNWRAPSTOP stops the writer from "wrapping" to the beginning of
   1311  * the disk when there are no clean segments higher than the current
   1312  * write offset.  This is primarily used for diagnostic and analysis
   1313  * purposes.  If an argument of 1 is given, this call waits until the
   1314  * writer stops.
   1315  *
   1316  * LFCNWRAPSTATUS retrieves the status of the wrapping restriction
   1317  * (going or stopping).
   1318  *
   1319  * LFCNWRAPPASS allows the segment selector to pass to the beginning
   1320  * of the disk, once, while wrapping is suspended.
   1321  *
   1322  * LFCNWRAPGO releases the restriction.
   1323  */
   1324 #define LFCNWRAPSTOP	_FCNR_FSPRIV('L',  9, int)
   1325 #define LFCNWRAPGO	_FCNR_FSPRIV('L', 10, int)
   1326 #define LFCNWRAPPASS	_FCNR_FSPRIV('L', 12, int)
   1327 # define LFS_WRAP_GOING	  0x0
   1328 # define LFS_WRAP_WAITING 0x1
   1329 #define LFCNWRAPSTATUS	_FCNW_FSPRIV('L', 13, int)
   1330 
   1331 /*
   1332  * Retrieve the file handle of the index file.  It is preferable for the
   1333  * cleaner to hold the file handle of the index file rather than the file
   1334  * system root because the latter would prevent a clean unmount.
   1335  */
   1336 struct lfs_fhandle {
   1337 	char space[28];	/* FHANDLE_SIZE_COMPAT (but used from userland too) */
   1338 };
   1339 #define LFCNIFILEFH	_FCNW_FSPRIV('L', 11, struct lfs_fhandle)
   1340 
   1341 /*
   1342  * Fcntls to take the place of the historical lfs syscalls.
   1343  *
   1344  * LFCNSEGWAIT waits for this file system to advance to a new segment.
   1345  *
   1346  * LFCNSEGWAITALL waits for *any* LFS to advance to a new segment.
   1347  *
   1348  * LFCNCLEANERINFO retrieves the most current cleaner information
   1349  * from the kernel.
   1350  *
   1351  * LFCNSEGUSE retrieves the current SEGUSE structure for segments
   1352  * beginning at _start_ and ending at _start_ + _len_ - 1.
   1353  *
   1354  * LFCNBMAPV takes a list of blocks, parsed out of the partial-segment
   1355  * headers, and returns those that have not been relocated.
   1356  *
   1357  * LFCNMARKV relocated the specified blocks, writing them at the head
   1358  * of the log.
   1359  *
   1360  * LFCNRECLAIM forces a checkpoint with the intention of reclaiming
   1361  * empty segments.  The write statistics show the amount of data
   1362  * written to disk to accomplish this.
   1363  */
   1364 #define LFS_SEGUSE_MAXCNT	65536	/* Max entry count for LFCNSEGUSE */
   1365 struct lfs_seguse_array {
   1366 	int start;	/* Start segment */
   1367 	size_t len;	/* Length of the array */
   1368 	SEGUSE *seguse; /* Out: array of seguse */
   1369 };
   1370 #define LFS_MARKV_MAXBLKCNT	65536	/* Max block count for lfs_markv() */
   1371 struct lfs_fcntl_markv {
   1372 	BLOCK_INFO *blkiov;	/* blocks to relocate */
   1373 	int blkcnt;		/* number of blocks (limited to 65536) */
   1374 };
   1375 #define LFCNSEGWAITALL	_FCNR_FSPRIV ('L', 14, struct timeval)
   1376 #define LFCNSEGWAIT	_FCNR_FSPRIV ('L', 15, struct timeval)
   1377 #define LFCNCLEANERINFO _FCNW_FSPRIV ('L', 22, CLEANERINFO64)
   1378 #define LFCNSEGUSE      _FCNRW_FSPRIV('L', 23, struct lfs_seguse_array)
   1379 #define LFCNBMAPV	_FCNRW_FSPRIV('L', 16, struct lfs_fcntl_markv)
   1380 #define LFCNMARKV	_FCNRW_FSPRIV('L', 17, struct lfs_fcntl_markv)
   1381 #define LFCNRECLAIM	_FCNW_FSPRIV ('L',  4, struct lfs_write_stats)
   1382 
   1383 /*
   1384  * File defragmentation.
   1385  *
   1386  * LFCNFILESTATS returns fragmentation information about the direct blocks
   1387  * held by the inodes numbered between _ino_ and _ino_ + _len_ - 1.
   1388  * Two metrics are returned: the total number of discontinuities; and the
   1389  * sum of the length of the discontinuities, reported in frags.
   1390  *
   1391  * LFCNREWRITEFILE rewrites the specified files sequentially,
   1392  * restoring them to a state with as few discontinuities as possible.
   1393  *
   1394  * LFCNSCRAMBLE rewrites every other block of the specified files,
   1395  * creating as many gaps as possible.  This is primarily used for
   1396  * testing.
   1397  */
   1398 #define LFS_FILESTATS_MAXCNT	65536	/* Max entry count for LFCNFILESTATS */
   1399 struct lfs_filestats {
   1400 	ino_t ino;	/* Inode number */
   1401 	off_t nblk;	/* Block count */
   1402 	int dc_count;	/* Count of discontinuities (gaps) */
   1403 	daddr_t dc_sum;	/* Sum of absolute values of gap lengths */
   1404 };
   1405 struct lfs_filestat_req {
   1406 	ino_t ino;	/* Starting inode number */
   1407 	int len;	/* Number of inodes to report */
   1408 	struct lfs_filestats *fss; /* Out: file statistics */
   1409 };
   1410 #define LFCNFILESTATS	_FCNRW_FSPRIV('L', 18, struct lfs_filestat_req)
   1411 
   1412 #define LFS_REWRITE_MAXCNT	65536	/* Max entry count for LFCNREWRITE* */
   1413 struct lfs_inode_array {
   1414 	size_t len;	/* Length of array */
   1415 	ino_t *inodes;	/* Array of inode numbers */
   1416 	struct lfs_write_stats stats; /* Out: frags written */
   1417 };
   1418 #define LFCNREWRITEFILE	_FCNRW_FSPRIV('L', 19, struct lfs_inode_array)
   1419 #define LFCNSCRAMBLE	_FCNRW_FSPRIV('L', 20, struct lfs_inode_array)
   1420 
   1421 /*
   1422  * Segment rewriting.
   1423  *
   1424  * LFCNREWRITESEGS rewrites the live data still contained within the
   1425  * specified segments.  It is intended to replace the
   1426  * read/lfs_bmapv/lfs_bmapv combination historically used by the cleaner.
   1427  */
   1428 struct lfs_segnum_array {
   1429 	size_t len;	/* Length of array */
   1430 	int *segments;	/* Array of segment numbers */
   1431 	struct lfs_write_stats stats; /* Out: frags written */
   1432 };
   1433 #define LFCNREWRITESEGS	_FCNRW_FSPRIV('L', 21, struct lfs_segnum_array)
   1434 
   1435 /*
   1436  * In-kernel cleaner.
   1437  *
   1438  * LFCNAUTOCLEAN directs the in-kernel cleaner process.  Three
   1439  * parameters are available: cleaning mode, priority threshold, and
   1440  * cleaning target.
   1441  *
   1442  * struct lfs_autoclean_params {
   1443  *      size_t size;  // Size of this structure
   1444  * #define LFS_CLEANMODE_NONE   0x00 // Auto-cleaner off
   1445  * #define LFS_CLEANMODE_GREEDY 0x01 // Greedy selector
   1446  * #define LFS_CLEANMODE_CB     0x02 // Cost-benefit selector
   1447  * 	int mode;     // Which of the autoclean modes to use
   1448  *	int thresh;   // Low cleaning threshold to use
   1449  *	int target;   // How many blocks to rewrite before checkpointing
   1450  * };
   1451  *
   1452  * The cleaning mode, or strategy, can be given as any of greedy;
   1453  * cost/benefit using clock time for the time metric; or NONE,
   1454  * indicating that the in-kernel cleaner should not be used for this
   1455  * file system.  The default is NONE.
   1456  *
   1457  * The priority threshold determines when to stop cleaning.  A
   1458  * threshold of zero indicates that the cleaner should never pause
   1459  * from cleaning until there are no segments with dead blocks; a
   1460  * negative value means the kernel will choose the threshold.
   1461  *
   1462  * The "target" parameter describes how many blocks should be written
   1463  * before a checkpoint.  Too low a value will result in the creation
   1464  * of many short-lived index file blocks, which might make the cleaner
   1465  * lose ground; but too high a value might result in the file system
   1466  * becoming short on clean segments, so that other processes must wait
   1467  * for a synchronous checkpoint.  A negative value means that the
   1468  * kernel will choose the target.
   1469  */
   1470 #define LFCNAUTOCLEAN	_FCNR_FSPRIV('L', 24, struct lfs_autoclean_params)
   1471 
   1472 
   1473 /* Debug segment lock */
   1474 #ifdef notyet
   1475 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs))
   1476 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs))
   1477 # define ASSERT_DUNNO_SEGLOCK(fs)
   1478 # define ASSERT_MAYBE_SEGLOCK(fs)
   1479 #else /* !notyet */
   1480 # define ASSERT_DUNNO_SEGLOCK(fs) \
   1481 	DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \
   1482 		LFS_SEGLOCK_HELD(fs)))
   1483 # define ASSERT_SEGLOCK(fs) do {					\
   1484 	if (!LFS_SEGLOCK_HELD(fs)) {					\
   1485 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \
   1486 	}								\
   1487 } while(0)
   1488 # define ASSERT_NO_SEGLOCK(fs) do {					\
   1489 	if (LFS_SEGLOCK_HELD(fs)) {					\
   1490 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \
   1491 	}								\
   1492 } while(0)
   1493 # define ASSERT_MAYBE_SEGLOCK(x)
   1494 #endif /* !notyet */
   1495 
   1496 /*
   1497  * Arguments to mount LFS filesystems
   1498  */
   1499 struct ulfs_args {
   1500 	char	*fspec;			/* block special device to mount */
   1501 };
   1502 
   1503 __BEGIN_DECLS
   1504 void lfs_itimes(struct inode *, const struct timespec *,
   1505     const struct timespec *, const struct timespec *);
   1506 __END_DECLS
   1507 
   1508 #endif /* !_UFS_LFS_LFS_H_ */
   1509