Home | History | Annotate | Line # | Download | only in lfs
lfs.h revision 1.147
      1 /*	$NetBSD: lfs.h,v 1.147 2013/06/08 02:15:12 dholland Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Konrad E. Schroder <perseant (at) hhhh.org>.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 /*-
     32  * Copyright (c) 1991, 1993
     33  *	The Regents of the University of California.  All rights reserved.
     34  *
     35  * Redistribution and use in source and binary forms, with or without
     36  * modification, are permitted provided that the following conditions
     37  * are met:
     38  * 1. Redistributions of source code must retain the above copyright
     39  *    notice, this list of conditions and the following disclaimer.
     40  * 2. Redistributions in binary form must reproduce the above copyright
     41  *    notice, this list of conditions and the following disclaimer in the
     42  *    documentation and/or other materials provided with the distribution.
     43  * 3. Neither the name of the University nor the names of its contributors
     44  *    may be used to endorse or promote products derived from this software
     45  *    without specific prior written permission.
     46  *
     47  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     57  * SUCH DAMAGE.
     58  *
     59  *	@(#)lfs.h	8.9 (Berkeley) 5/8/95
     60  */
     61 
     62 #ifndef _UFS_LFS_LFS_H_
     63 #define _UFS_LFS_LFS_H_
     64 
     65 #include <sys/rwlock.h>
     66 #include <sys/mutex.h>
     67 #include <sys/queue.h>
     68 #include <sys/condvar.h>
     69 #include <sys/mount.h>
     70 #include <sys/pool.h>
     71 
     72 /*
     73  * Compile-time options for LFS.
     74  */
     75 #define LFS_IFIND_RETRIES  16
     76 #define LFS_LOGLENGTH      1024 /* size of debugging log */
     77 #define LFS_MAX_ACTIVE	   10	/* Dirty segments before ckp forced */
     78 
     79 /*
     80  * Fixed filesystem layout parameters
     81  */
     82 #define	LFS_LABELPAD	8192		/* LFS label size */
     83 #define	LFS_SBPAD	8192		/* LFS superblock size */
     84 
     85 #define	LFS_UNUSED_INUM	0		/* 0: out of band inode number */
     86 #define	LFS_IFILE_INUM	1		/* 1: IFILE inode number */
     87 					/* 2: Root inode number */
     88 #define	LFS_LOSTFOUNDINO 3		/* 3: lost+found inode number */
     89 #define	LFS_FIRST_INUM	4		/* 4: first free inode number */
     90 
     91 /*
     92  * The root inode is the root of the file system.  Inode 0 can't be used for
     93  * normal purposes and historically bad blocks were linked to inode 1, thus
     94  * the root inode is 2.  (Inode 1 is no longer used for this purpose, however
     95  * numerous dump tapes make this assumption, so we are stuck with it).
     96  */
     97 #define	ULFS_ROOTINO	((ino_t)2)
     98 
     99 /*
    100  * The Whiteout inode# is a dummy non-zero inode number which will
    101  * never be allocated to a real file.  It is used as a place holder
    102  * in the directory entry which has been tagged as a LFS_DT_WHT entry.
    103  * See the comments about ULFS_ROOTINO above.
    104  */
    105 #define	ULFS_WINO	((ino_t)1)
    106 
    107 
    108 #define	LFS_V1_SUMMARY_SIZE	512     /* V1 fixed summary size */
    109 #define	LFS_DFL_SUMMARY_SIZE	512	/* Default summary size */
    110 
    111 #define LFS_MAX_DADDR	0x7fffffff	/* Highest addressable fsb */
    112 
    113 #define LFS_MAXNAMLEN	255		/* maximum name length in a dir */
    114 
    115 #define ULFS_NXADDR	2
    116 #define	ULFS_NDADDR	12		/* Direct addresses in inode. */
    117 #define	ULFS_NIADDR	3		/* Indirect addresses in inode. */
    118 
    119 /*
    120  * Adjustable filesystem parameters
    121  */
    122 #define MIN_FREE_SEGS	20
    123 #define MIN_RESV_SEGS	15
    124 #ifndef LFS_ATIME_IFILE
    125 # define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */
    126 #endif
    127 #define LFS_MARKV_MAXBLKCNT	65536	/* Max block count for lfs_markv() */
    128 
    129 /* Misc. definitions */
    130 #define BW_CLEAN	1		/* Flag for lfs_bwrite_ext() */
    131 #define PG_DELWRI	PG_PAGER1	/* Local def for delayed pageout */
    132 
    133 /* Resource limits */
    134 #define	LFS_MAX_RESOURCE(x, u)	(((x) >> 2) - 10 * (u))
    135 #define	LFS_WAIT_RESOURCE(x, u)	(((x) >> 1) - ((x) >> 3) - 10 * (u))
    136 #define	LFS_INVERSE_MAX_RESOURCE(x, u)	(((x) + 10 * (u)) << 2)
    137 #define LFS_MAX_BUFS	    LFS_MAX_RESOURCE(nbuf, 1)
    138 #define LFS_WAIT_BUFS	    LFS_WAIT_RESOURCE(nbuf, 1)
    139 #define LFS_INVERSE_MAX_BUFS(n)	LFS_INVERSE_MAX_RESOURCE(n, 1)
    140 #define LFS_MAX_BYTES	    LFS_MAX_RESOURCE(bufmem_lowater, PAGE_SIZE)
    141 #define LFS_INVERSE_MAX_BYTES(n) LFS_INVERSE_MAX_RESOURCE(n, PAGE_SIZE)
    142 #define LFS_WAIT_BYTES	    LFS_WAIT_RESOURCE(bufmem_lowater, PAGE_SIZE)
    143 #define LFS_MAX_DIROP	    ((desiredvnodes >> 2) + (desiredvnodes >> 3))
    144 #define SIZEOF_DIROP(fs)	(2 * ((fs)->lfs_bsize + LFS_DINODE1_SIZE))
    145 #define LFS_MAX_FSDIROP(fs)						\
    146 	((fs)->lfs_nclean <= (fs)->lfs_resvseg ? 0 :			\
    147 	 (((fs)->lfs_nclean - (fs)->lfs_resvseg) * (fs)->lfs_ssize) /	\
    148           (2 * SIZEOF_DIROP(fs)))
    149 #define LFS_MAX_PAGES	lfs_max_pages()
    150 #define LFS_WAIT_PAGES	lfs_wait_pages()
    151 #define LFS_BUFWAIT	    2	/* How long to wait if over *_WAIT_* */
    152 
    153 #ifdef _KERNEL
    154 int lfs_wait_pages(void);
    155 int lfs_max_pages(void);
    156 #endif /* _KERNEL */
    157 
    158 /* How starved can we be before we start holding back page writes */
    159 #define LFS_STARVED_FOR_SEGS(fs) ((fs)->lfs_nclean < (fs)->lfs_resvseg)
    160 
    161 /*
    162  * Reserved blocks for lfs_malloc
    163  */
    164 
    165 /* Structure to keep reserved blocks */
    166 typedef struct lfs_res_blk {
    167 	void *p;
    168 	LIST_ENTRY(lfs_res_blk) res;
    169 	int size;
    170 	char inuse;
    171 } res_t;
    172 
    173 /* Types for lfs_newbuf and lfs_malloc */
    174 #define LFS_NB_UNKNOWN -1
    175 #define LFS_NB_SUMMARY	0
    176 #define LFS_NB_SBLOCK	1
    177 #define LFS_NB_IBLOCK	2
    178 #define LFS_NB_CLUSTER	3
    179 #define LFS_NB_CLEAN	4
    180 #define LFS_NB_BLKIOV	5
    181 #define LFS_NB_COUNT	6 /* always last */
    182 
    183 /* Number of reserved memory blocks of each type */
    184 #define LFS_N_SUMMARIES 2
    185 #define LFS_N_SBLOCKS	1   /* Always 1, to throttle superblock writes */
    186 #define LFS_N_IBLOCKS	16  /* In theory ssize/bsize; in practice around 2 */
    187 #define LFS_N_CLUSTERS	16  /* In theory ssize/MAXPHYS */
    188 #define LFS_N_CLEAN	0
    189 #define LFS_N_BLKIOV	1
    190 
    191 /* Total count of "large" (non-pool) types */
    192 #define LFS_N_TOTAL (LFS_N_SUMMARIES + LFS_N_SBLOCKS + LFS_N_IBLOCKS +	\
    193 		     LFS_N_CLUSTERS + LFS_N_CLEAN + LFS_N_BLKIOV)
    194 
    195 /* Counts for pool types */
    196 #define LFS_N_CL	LFS_N_CLUSTERS
    197 #define LFS_N_BPP	2
    198 #define LFS_N_SEG	2
    199 
    200 /*
    201  * Directories
    202  */
    203 
    204 /*
    205  * A directory consists of some number of blocks of LFS_DIRBLKSIZ
    206  * bytes, where LFS_DIRBLKSIZ is chosen such that it can be transferred
    207  * to disk in a single atomic operation (e.g. 512 bytes on most machines).
    208  *
    209  * Each LFS_DIRBLKSIZ byte block contains some number of directory entry
    210  * structures, which are of variable length.  Each directory entry has
    211  * a struct lfs_direct at the front of it, containing its inode number,
    212  * the length of the entry, and the length of the name contained in
    213  * the entry.  These are followed by the name padded to a 4 byte boundary.
    214  * All names are guaranteed null terminated.
    215  * The maximum length of a name in a directory is LFS_MAXNAMLEN.
    216  *
    217  * The macro DIRSIZ(fmt, dp) gives the amount of space required to represent
    218  * a directory entry.  Free space in a directory is represented by
    219  * entries which have dp->d_reclen > DIRSIZ(fmt, dp).  All LFS_DIRBLKSIZ bytes
    220  * in a directory block are claimed by the directory entries.  This
    221  * usually results in the last entry in a directory having a large
    222  * dp->d_reclen.  When entries are deleted from a directory, the
    223  * space is returned to the previous entry in the same directory
    224  * block by increasing its dp->d_reclen.  If the first entry of
    225  * a directory block is free, then its dp->d_ino is set to 0.
    226  * Entries other than the first in a directory do not normally have
    227  * dp->d_ino set to 0.
    228  */
    229 
    230 /*
    231  * Directory block size.
    232  */
    233 #undef	LFS_DIRBLKSIZ
    234 #define	LFS_DIRBLKSIZ	DEV_BSIZE
    235 
    236 /*
    237  * Convert between stat structure types and directory types.
    238  */
    239 #define	LFS_IFTODT(mode)	(((mode) & 0170000) >> 12)
    240 #define	LFS_DTTOIF(dirtype)	((dirtype) << 12)
    241 
    242 /*
    243  * The LFS_DIRSIZ macro gives the minimum record length which will hold
    244  * the directory entry.  This requires the amount of space in struct lfs_direct
    245  * without the d_name field, plus enough space for the name with a terminating
    246  * null byte (dp->d_namlen+1), rounded up to a 4 byte boundary.
    247  */
    248 #define	LFS_DIRECTSIZ(namlen) \
    249 	((sizeof(struct lfs_direct) - (LFS_MAXNAMLEN+1)) + (((namlen)+1 + 3) &~ 3))
    250 
    251 #if (BYTE_ORDER == LITTLE_ENDIAN)
    252 #define LFS_DIRSIZ(oldfmt, dp, needswap)	\
    253     (((oldfmt) && !(needswap)) ?		\
    254     LFS_DIRECTSIZ((dp)->d_type) : LFS_DIRECTSIZ((dp)->d_namlen))
    255 #else
    256 #define LFS_DIRSIZ(oldfmt, dp, needswap)	\
    257     (((oldfmt) && (needswap)) ?			\
    258     LFS_DIRECTSIZ((dp)->d_type) : LFS_DIRECTSIZ((dp)->d_namlen))
    259 #endif
    260 
    261 /* Constants for the first argument of LFS_DIRSIZ */
    262 #define LFS_OLDDIRFMT	1
    263 #define LFS_NEWDIRFMT	0
    264 
    265 /*
    266  * Theoretically, directories can be more than 2Gb in length; however, in
    267  * practice this seems unlikely. So, we define the type doff_t as a 32-bit
    268  * quantity to keep down the cost of doing lookup on a 32-bit machine.
    269  */
    270 #define	doff_t		int32_t
    271 #define	lfs_doff_t	int32_t
    272 #define	MAXDIRSIZE	(0x7fffffff)
    273 
    274 #define	LFS_MAXNAMLEN	255
    275 
    276 /*
    277  * File types for d_type
    278  */
    279 #define	LFS_DT_UNKNOWN	 0
    280 #define	LFS_DT_FIFO	 1
    281 #define	LFS_DT_CHR	 2
    282 #define	LFS_DT_DIR	 4
    283 #define	LFS_DT_BLK	 6
    284 #define	LFS_DT_REG	 8
    285 #define	LFS_DT_LNK	10
    286 #define	LFS_DT_SOCK	12
    287 #define	LFS_DT_WHT	14
    288 
    289 /*
    290  * (See notes in ulfs_dir.h)
    291  */
    292 #define d_ino d_fileno
    293 struct lfs_direct {
    294 	u_int32_t d_fileno;		/* inode number of entry */
    295 	u_int16_t d_reclen;		/* length of this record */
    296 	u_int8_t  d_type; 		/* file type, see below */
    297 	u_int8_t  d_namlen;		/* length of string in d_name */
    298 	char	  d_name[LFS_MAXNAMLEN + 1];/* name with length <= LFS_MAXNAMLEN */
    299 };
    300 
    301 /*
    302  * Template for manipulating directories.  Should use struct lfs_direct's,
    303  * but the name field is LFS_MAXNAMLEN - 1, and this just won't do.
    304  */
    305 struct lfs_dirtemplate {
    306 	u_int32_t	dot_ino;
    307 	int16_t		dot_reclen;
    308 	u_int8_t	dot_type;
    309 	u_int8_t	dot_namlen;
    310 	char		dot_name[4];	/* must be multiple of 4 */
    311 	u_int32_t	dotdot_ino;
    312 	int16_t		dotdot_reclen;
    313 	u_int8_t	dotdot_type;
    314 	u_int8_t	dotdot_namlen;
    315 	char		dotdot_name[4];	/* ditto */
    316 };
    317 
    318 /*
    319  * This is the old format of directories, sans type element.
    320  */
    321 struct lfs_odirtemplate {
    322 	u_int32_t	dot_ino;
    323 	int16_t		dot_reclen;
    324 	u_int16_t	dot_namlen;
    325 	char		dot_name[4];	/* must be multiple of 4 */
    326 	u_int32_t	dotdot_ino;
    327 	int16_t		dotdot_reclen;
    328 	u_int16_t	dotdot_namlen;
    329 	char		dotdot_name[4];	/* ditto */
    330 };
    331 
    332 /*
    333  * Inodes
    334  */
    335 
    336 /*
    337  * A dinode contains all the meta-data associated with a ULFS file.
    338  * This structure defines the on-disk format of a dinode. Since
    339  * this structure describes an on-disk structure, all its fields
    340  * are defined by types with precise widths.
    341  */
    342 
    343 struct ulfs1_dinode {
    344 	u_int16_t	di_mode;	/*   0: IFMT, permissions; see below. */
    345 	int16_t		di_nlink;	/*   2: File link count. */
    346 	union {
    347 		u_int16_t oldids[2];	/*   4: Ffs: old user and group ids. */
    348 		u_int32_t inumber;	/*   4: Lfs: inode number. */
    349 	} di_u;
    350 	u_int64_t	di_size;	/*   8: File byte count. */
    351 	int32_t		di_atime;	/*  16: Last access time. */
    352 	int32_t		di_atimensec;	/*  20: Last access time. */
    353 	int32_t		di_mtime;	/*  24: Last modified time. */
    354 	int32_t		di_mtimensec;	/*  28: Last modified time. */
    355 	int32_t		di_ctime;	/*  32: Last inode change time. */
    356 	int32_t		di_ctimensec;	/*  36: Last inode change time. */
    357 	int32_t		di_db[ULFS_NDADDR]; /*  40: Direct disk blocks. */
    358 	int32_t		di_ib[ULFS_NIADDR]; /*  88: Indirect disk blocks. */
    359 	u_int32_t	di_flags;	/* 100: Status flags (chflags). */
    360 	u_int32_t	di_blocks;	/* 104: Blocks actually held. */
    361 	int32_t		di_gen;		/* 108: Generation number. */
    362 	u_int32_t	di_uid;		/* 112: File owner. */
    363 	u_int32_t	di_gid;		/* 116: File group. */
    364 	u_int64_t	di_modrev;	/* 120: i_modrev for NFSv4 */
    365 };
    366 
    367 struct ulfs2_dinode {
    368 	u_int16_t	di_mode;	/*   0: IFMT, permissions; see below. */
    369 	int16_t		di_nlink;	/*   2: File link count. */
    370 	u_int32_t	di_uid;		/*   4: File owner. */
    371 	u_int32_t	di_gid;		/*   8: File group. */
    372 	u_int32_t	di_blksize;	/*  12: Inode blocksize. */
    373 	u_int64_t	di_size;	/*  16: File byte count. */
    374 	u_int64_t	di_blocks;	/*  24: Bytes actually held. */
    375 	int64_t		di_atime;	/*  32: Last access time. */
    376 	int64_t		di_mtime;	/*  40: Last modified time. */
    377 	int64_t		di_ctime;	/*  48: Last inode change time. */
    378 	int64_t		di_birthtime;	/*  56: Inode creation time. */
    379 	int32_t		di_mtimensec;	/*  64: Last modified time. */
    380 	int32_t		di_atimensec;	/*  68: Last access time. */
    381 	int32_t		di_ctimensec;	/*  72: Last inode change time. */
    382 	int32_t		di_birthnsec;	/*  76: Inode creation time. */
    383 	int32_t		di_gen;		/*  80: Generation number. */
    384 	u_int32_t	di_kernflags;	/*  84: Kernel flags. */
    385 	u_int32_t	di_flags;	/*  88: Status flags (chflags). */
    386 	int32_t		di_extsize;	/*  92: External attributes block. */
    387 	int64_t		di_extb[ULFS_NXADDR];/* 96: External attributes block. */
    388 	int64_t		di_db[ULFS_NDADDR]; /* 112: Direct disk blocks. */
    389 	int64_t		di_ib[ULFS_NIADDR]; /* 208: Indirect disk blocks. */
    390 	u_int64_t	di_modrev;	/* 232: i_modrev for NFSv4 */
    391 	int64_t		di_spare[2];	/* 240: Reserved; currently unused */
    392 };
    393 
    394 /*
    395  * The di_db fields may be overlaid with other information for
    396  * file types that do not have associated disk storage. Block
    397  * and character devices overlay the first data block with their
    398  * dev_t value. Short symbolic links place their path in the
    399  * di_db area.
    400  */
    401 #define	di_inumber	di_u.inumber
    402 #define	di_ogid		di_u.oldids[1]
    403 #define	di_ouid		di_u.oldids[0]
    404 #define	di_rdev		di_db[0]
    405 
    406 /* Size of the on-disk inode. */
    407 #define	LFS_DINODE1_SIZE	(sizeof(struct ulfs1_dinode))	/* 128 */
    408 #define	LFS_DINODE2_SIZE	(sizeof(struct ulfs2_dinode))
    409 
    410 /* File types, found in the upper bits of di_mode. */
    411 #define	LFS_IFMT	0170000		/* Mask of file type. */
    412 #define	LFS_IFIFO	0010000		/* Named pipe (fifo). */
    413 #define	LFS_IFCHR	0020000		/* Character device. */
    414 #define	LFS_IFDIR	0040000		/* Directory file. */
    415 #define	LFS_IFBLK	0060000		/* Block device. */
    416 #define	LFS_IFREG	0100000		/* Regular file. */
    417 #define	LFS_IFLNK	0120000		/* Symbolic link. */
    418 #define	LFS_IFSOCK	0140000		/* UNIX domain socket. */
    419 #define	LFS_IFWHT	0160000		/* Whiteout. */
    420 
    421 /*
    422  * Maximum length of a symlink that can be stored within the inode.
    423  */
    424 #define ULFS1_MAXSYMLINKLEN	((ULFS_NDADDR + ULFS_NIADDR) * sizeof(int32_t))
    425 #define ULFS2_MAXSYMLINKLEN	((ULFS_NDADDR + ULFS_NIADDR) * sizeof(int64_t))
    426 
    427 #define ULFS_MAXSYMLINKLEN(ip) \
    428 	((ip)->i_ump->um_fstype == ULFS1) ? \
    429 	ULFS1_MAXSYMLINKLEN : ULFS2_MAXSYMLINKLEN
    430 
    431 /*
    432  * "struct buf" associated definitions
    433  */
    434 
    435 /* Unassigned disk addresses. */
    436 #define	UNASSIGNED	-1
    437 #define UNWRITTEN	-2
    438 
    439 /* Unused logical block number */
    440 #define LFS_UNUSED_LBN	-1
    441 
    442 /* Determine if a buffer belongs to the ifile */
    443 #define IS_IFILE(bp)	(VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM)
    444 
    445 # define LFS_LOCK_BUF(bp) do {						\
    446 	if (((bp)->b_flags & B_LOCKED) == 0 && bp->b_iodone == NULL) {	\
    447 		mutex_enter(&lfs_lock);					\
    448 		++locked_queue_count;					\
    449 		locked_queue_bytes += bp->b_bufsize;			\
    450 		mutex_exit(&lfs_lock);					\
    451 	}								\
    452 	(bp)->b_flags |= B_LOCKED;					\
    453 } while (0)
    454 
    455 # define LFS_UNLOCK_BUF(bp) do {					\
    456 	if (((bp)->b_flags & B_LOCKED) != 0 && bp->b_iodone == NULL) {	\
    457 		mutex_enter(&lfs_lock);					\
    458 		--locked_queue_count;					\
    459 		locked_queue_bytes -= bp->b_bufsize;			\
    460 		if (locked_queue_count < LFS_WAIT_BUFS &&		\
    461 		    locked_queue_bytes < LFS_WAIT_BYTES)		\
    462 			cv_broadcast(&locked_queue_cv);			\
    463 		mutex_exit(&lfs_lock);					\
    464 	}								\
    465 	(bp)->b_flags &= ~B_LOCKED;					\
    466 } while (0)
    467 
    468 #ifdef _KERNEL
    469 
    470 extern u_long bufmem_lowater, bufmem_hiwater; /* XXX */
    471 
    472 # define LFS_IS_MALLOC_BUF(bp) ((bp)->b_iodone == lfs_callback)
    473 
    474 # ifdef DEBUG
    475 #  define LFS_DEBUG_COUNTLOCKED(m) do {					\
    476 	if (lfs_debug_log_subsys[DLOG_LLIST]) {				\
    477 		lfs_countlocked(&locked_queue_count, &locked_queue_bytes, (m)); \
    478 		cv_broadcast(&locked_queue_cv);				\
    479 	}								\
    480 } while (0)
    481 # else
    482 #  define LFS_DEBUG_COUNTLOCKED(m)
    483 # endif
    484 
    485 /* log for debugging writes to the Ifile */
    486 # ifdef DEBUG
    487 struct lfs_log_entry {
    488 	const char *op;
    489 	const char *file;
    490 	int pid;
    491 	int line;
    492 	daddr_t block;
    493 	unsigned long flags;
    494 };
    495 extern int lfs_lognum;
    496 extern struct lfs_log_entry lfs_log[LFS_LOGLENGTH];
    497 #  define LFS_BWRITE_LOG(bp) lfs_bwrite_log((bp), __FILE__, __LINE__)
    498 #  define LFS_ENTER_LOG(theop, thefile, theline, lbn, theflags, thepid) do {\
    499 	int _s;								\
    500 									\
    501 	mutex_enter(&lfs_lock);						\
    502 	_s = splbio();							\
    503 	lfs_log[lfs_lognum].op = theop;					\
    504 	lfs_log[lfs_lognum].file = thefile;				\
    505 	lfs_log[lfs_lognum].line = (theline);				\
    506 	lfs_log[lfs_lognum].pid = (thepid);				\
    507 	lfs_log[lfs_lognum].block = (lbn);				\
    508 	lfs_log[lfs_lognum].flags = (theflags);				\
    509 	lfs_lognum = (lfs_lognum + 1) % LFS_LOGLENGTH;			\
    510 	splx(_s);							\
    511 	mutex_exit(&lfs_lock);						\
    512 } while (0)
    513 
    514 #  define LFS_BCLEAN_LOG(fs, bp) do {					\
    515 	if ((bp)->b_vp == (fs)->lfs_ivnode)				\
    516 		LFS_ENTER_LOG("clear", __FILE__, __LINE__,		\
    517 			      bp->b_lblkno, bp->b_flags, curproc->p_pid);\
    518 } while (0)
    519 
    520 /* Must match list in lfs_vfsops.c ! */
    521 #  define DLOG_RF     0  /* roll forward */
    522 #  define DLOG_ALLOC  1  /* inode alloc */
    523 #  define DLOG_AVAIL  2  /* lfs_{,r,f}avail */
    524 #  define DLOG_FLUSH  3  /* flush */
    525 #  define DLOG_LLIST  4  /* locked list accounting */
    526 #  define DLOG_WVNODE 5  /* vflush/writevnodes verbose */
    527 #  define DLOG_VNODE  6  /* vflush/writevnodes */
    528 #  define DLOG_SEG    7  /* segwrite */
    529 #  define DLOG_SU     8  /* seguse accounting */
    530 #  define DLOG_CLEAN  9  /* cleaner routines */
    531 #  define DLOG_MOUNT  10 /* mount/unmount */
    532 #  define DLOG_PAGE   11 /* putpages/gop_write */
    533 #  define DLOG_DIROP  12 /* dirop accounting */
    534 #  define DLOG_MALLOC 13 /* lfs_malloc accounting */
    535 #  define DLOG_MAX    14 /* The terminator */
    536 #  define DLOG(a) lfs_debug_log a
    537 # else /* ! DEBUG */
    538 #  define LFS_BCLEAN_LOG(fs, bp)
    539 #  define LFS_BWRITE_LOG(bp)		VOP_BWRITE((bp)->b_vp, (bp))
    540 #  define DLOG(a)
    541 # endif /* ! DEBUG */
    542 #else /* ! _KERNEL */
    543 # define LFS_BWRITE_LOG(bp)		VOP_BWRITE((bp))
    544 #endif /* _KERNEL */
    545 
    546 #ifdef _KERNEL
    547 /* This overlays the fid structure (see fstypes.h). */
    548 struct ulfs_ufid {
    549 	u_int16_t ufid_len;	/* Length of structure. */
    550 	u_int16_t ufid_pad;	/* Force 32-bit alignment. */
    551 	u_int32_t ufid_ino;	/* File number (ino). */
    552 	int32_t	  ufid_gen;	/* Generation number. */
    553 };
    554 /* Filehandle structure for exported LFSes */
    555 struct lfid {
    556 	struct ulfs_ufid lfid_ufid;
    557 #define lfid_len lfid_ufid.ufid_len
    558 #define lfid_ino lfid_ufid.ufid_ino
    559 #define lfid_gen lfid_ufid.ufid_gen
    560 	uint32_t lfid_ident;
    561 };
    562 #endif /* _KERNEL */
    563 
    564 /*
    565  * "struct inode" associated definitions
    566  */
    567 
    568 /* Address calculations for metadata located in the inode */
    569 #define	S_INDIR(fs)	-ULFS_NDADDR
    570 #define	D_INDIR(fs)	(S_INDIR(fs) - NINDIR(fs) - 1)
    571 #define	T_INDIR(fs)	(D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1)
    572 
    573 /* For convenience */
    574 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING)
    575 
    576 #define LFS_SET_UINO(ip, flags) do {					\
    577 	if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED))	\
    578 		++(ip)->i_lfs->lfs_uinodes;				\
    579 	if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING))	\
    580 		++(ip)->i_lfs->lfs_uinodes;				\
    581 	if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED))	\
    582 		++(ip)->i_lfs->lfs_uinodes;				\
    583 	(ip)->i_flag |= (flags);					\
    584 } while (0)
    585 
    586 #define LFS_CLR_UINO(ip, flags) do {					\
    587 	if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED))	\
    588 		--(ip)->i_lfs->lfs_uinodes;				\
    589 	if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING))	\
    590 		--(ip)->i_lfs->lfs_uinodes;				\
    591 	if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED))	\
    592 		--(ip)->i_lfs->lfs_uinodes;				\
    593 	(ip)->i_flag &= ~(flags);					\
    594 	if ((ip)->i_lfs->lfs_uinodes < 0) {				\
    595 		panic("lfs_uinodes < 0");				\
    596 	}								\
    597 } while (0)
    598 
    599 #define LFS_ITIMES(ip, acc, mod, cre) \
    600 	while ((ip)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY)) \
    601 		lfs_itimes(ip, acc, mod, cre)
    602 
    603 /*
    604  * "struct vnode" associated definitions
    605  */
    606 
    607 /* Heuristic emptiness measure */
    608 #define VPISEMPTY(vp)	 (LIST_EMPTY(&(vp)->v_dirtyblkhd) && 		\
    609 			  !(vp->v_type == VREG && (vp)->v_iflag & VI_ONWORKLST) &&\
    610 			  VTOI(vp)->i_lfs_nbtree == 0)
    611 
    612 #define WRITEINPROG(vp) ((vp)->v_numoutput > 0 ||			\
    613 	(!LIST_EMPTY(&(vp)->v_dirtyblkhd) &&				\
    614 	 !(VTOI(vp)->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING))))
    615 
    616 
    617 /*
    618  * On-disk and in-memory checkpoint segment usage structure.
    619  */
    620 typedef struct segusage SEGUSE;
    621 struct segusage {
    622 	u_int32_t su_nbytes;		/* 0: number of live bytes */
    623 	u_int32_t su_olastmod;		/* 4: SEGUSE last modified timestamp */
    624 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
    625 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
    626 
    627 #define	SEGUSE_ACTIVE		0x01	/*  segment currently being written */
    628 #define	SEGUSE_DIRTY		0x02	/*  segment has data in it */
    629 #define	SEGUSE_SUPERBLOCK	0x04	/*  segment contains a superblock */
    630 #define SEGUSE_ERROR		0x08	/*  cleaner: do not clean segment */
    631 #define SEGUSE_EMPTY		0x10	/*  segment is empty */
    632 #define SEGUSE_INVAL		0x20	/*  segment is invalid */
    633 	u_int32_t su_flags;		/* 12: segment flags */
    634 	u_int64_t su_lastmod;		/* 16: last modified timestamp */
    635 };
    636 
    637 typedef struct segusage_v1 SEGUSE_V1;
    638 struct segusage_v1 {
    639 	u_int32_t su_nbytes;		/* 0: number of live bytes */
    640 	u_int32_t su_lastmod;		/* 4: SEGUSE last modified timestamp */
    641 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
    642 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
    643 	u_int32_t su_flags;		/* 12: segment flags  */
    644 };
    645 
    646 #define	SEGUPB(fs)	(fs->lfs_sepb)
    647 #define	SEGTABSIZE_SU(fs)						\
    648 	(((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb)
    649 
    650 #ifdef _KERNEL
    651 # define SHARE_IFLOCK(F) 						\
    652   do {									\
    653 	rw_enter(&(F)->lfs_iflock, RW_READER);				\
    654   } while(0)
    655 # define UNSHARE_IFLOCK(F)						\
    656   do {									\
    657 	rw_exit(&(F)->lfs_iflock);					\
    658   } while(0)
    659 #else /* ! _KERNEL */
    660 # define SHARE_IFLOCK(F)
    661 # define UNSHARE_IFLOCK(F)
    662 #endif /* ! _KERNEL */
    663 
    664 /* Read in the block with a specific segment usage entry from the ifile. */
    665 #define	LFS_SEGENTRY(SP, F, IN, BP) do {				\
    666 	int _e;								\
    667 	SHARE_IFLOCK(F);						\
    668 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
    669 	if ((_e = bread((F)->lfs_ivnode,				\
    670 	    ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz,			\
    671 	    (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0)			\
    672 		panic("lfs: ifile read: %d", _e);			\
    673 	if ((F)->lfs_version == 1)					\
    674 		(SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data +		\
    675 			((IN) & ((F)->lfs_sepb - 1)));			\
    676 	else								\
    677 		(SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb);	\
    678 	UNSHARE_IFLOCK(F);						\
    679 } while (0)
    680 
    681 #define LFS_WRITESEGENTRY(SP, F, IN, BP) do {				\
    682 	if ((SP)->su_nbytes == 0)					\
    683 		(SP)->su_flags |= SEGUSE_EMPTY;				\
    684 	else								\
    685 		(SP)->su_flags &= ~SEGUSE_EMPTY;			\
    686 	(F)->lfs_suflags[(F)->lfs_activesb][(IN)] = (SP)->su_flags;	\
    687 	LFS_BWRITE_LOG(BP);						\
    688 } while (0)
    689 
    690 /*
    691  * On-disk file information.  One per file with data blocks in the segment.
    692  */
    693 typedef struct finfo FINFO;
    694 struct finfo {
    695 	u_int32_t fi_nblocks;		/* number of blocks */
    696 	u_int32_t fi_version;		/* version number */
    697 	u_int32_t fi_ino;		/* inode number */
    698 	u_int32_t fi_lastlength;	/* length of last block in array */
    699 	int32_t	  fi_blocks[1];		/* array of logical block numbers */
    700 };
    701 /* sizeof FINFO except fi_blocks */
    702 #define	FINFOSIZE	(sizeof(FINFO) - sizeof(int32_t))
    703 
    704 /*
    705  * Index file inode entries.
    706  */
    707 typedef struct ifile IFILE;
    708 struct ifile {
    709 	u_int32_t if_version;		/* inode version number */
    710 #define	LFS_UNUSED_DADDR	0	/* out-of-band daddr */
    711 	int32_t	  if_daddr;		/* inode disk address */
    712 #define LFS_ORPHAN_NEXTFREE	(~(u_int32_t)0) /* indicate orphaned file */
    713 	u_int32_t if_nextfree;		/* next-unallocated inode */
    714 	u_int32_t if_atime_sec;		/* Last access time, seconds */
    715 	u_int32_t if_atime_nsec;	/* and nanoseconds */
    716 };
    717 
    718 typedef struct ifile_v1 IFILE_V1;
    719 struct ifile_v1 {
    720 	u_int32_t if_version;		/* inode version number */
    721 	int32_t	  if_daddr;		/* inode disk address */
    722 	u_int32_t if_nextfree;		/* next-unallocated inode */
    723 #if LFS_ATIME_IFILE
    724 	struct timespec if_atime;	/* Last access time */
    725 #endif
    726 };
    727 
    728 /*
    729  * LFSv1 compatibility code is not allowed to touch if_atime, since it
    730  * may not be mapped!
    731  */
    732 /* Read in the block with a specific inode from the ifile. */
    733 #define	LFS_IENTRY(IP, F, IN, BP) do {					\
    734 	int _e;								\
    735 	SHARE_IFLOCK(F);						\
    736 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
    737 	if ((_e = bread((F)->lfs_ivnode,				\
    738 	(IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz,	\
    739 	(F)->lfs_bsize, NOCRED, 0, &(BP))) != 0)			\
    740 		panic("lfs: ifile ino %d read %d", (int)(IN), _e);	\
    741 	if ((F)->lfs_version == 1)					\
    742 		(IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data +		\
    743 				 (IN) % (F)->lfs_ifpb);			\
    744 	else								\
    745 		(IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb;	\
    746 	UNSHARE_IFLOCK(F);						\
    747 } while (0)
    748 
    749 /*
    750  * Cleaner information structure.  This resides in the ifile and is used
    751  * to pass information from the kernel to the cleaner.
    752  */
    753 typedef struct _cleanerinfo {
    754 	u_int32_t clean;		/* number of clean segments */
    755 	u_int32_t dirty;		/* number of dirty segments */
    756 	int32_t   bfree;		/* disk blocks free */
    757 	int32_t	  avail;		/* disk blocks available */
    758 	u_int32_t free_head;		/* head of the inode free list */
    759 	u_int32_t free_tail;		/* tail of the inode free list */
    760 #define LFS_CLEANER_MUST_CLEAN	0x01
    761 	u_int32_t flags;		/* status word from the kernel */
    762 } CLEANERINFO;
    763 
    764 #define	CLEANSIZE_SU(fs)						\
    765 	((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift)
    766 
    767 /* Read in the block with the cleaner info from the ifile. */
    768 #define LFS_CLEANERINFO(CP, F, BP) do {					\
    769 	SHARE_IFLOCK(F);						\
    770 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
    771 	if (bread((F)->lfs_ivnode,					\
    772 	    (daddr_t)0, (F)->lfs_bsize, NOCRED, 0, &(BP)))		\
    773 		panic("lfs: ifile read");				\
    774 	(CP) = (CLEANERINFO *)(BP)->b_data;				\
    775 	UNSHARE_IFLOCK(F);						\
    776 } while (0)
    777 
    778 /*
    779  * Synchronize the Ifile cleaner info with current avail and bfree.
    780  */
    781 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do {		 	\
    782     mutex_enter(&lfs_lock);						\
    783     if ((w) || (cip)->bfree != (fs)->lfs_bfree ||		 	\
    784 	(cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail - 		\
    785 	(fs)->lfs_favail) {	 					\
    786 	(cip)->bfree = (fs)->lfs_bfree;				 	\
    787 	(cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail -		\
    788 		(fs)->lfs_favail;				 	\
    789 	if (((bp)->b_flags & B_GATHERED) == 0) {		 	\
    790 		(fs)->lfs_flags |= LFS_IFDIRTY;			 	\
    791 	}								\
    792 	mutex_exit(&lfs_lock);						\
    793 	(void) LFS_BWRITE_LOG(bp); /* Ifile */			 	\
    794     } else {							 	\
    795 	mutex_exit(&lfs_lock);						\
    796 	brelse(bp, 0);						 	\
    797     }									\
    798 } while (0)
    799 
    800 /*
    801  * Get the head of the inode free list.
    802  * Always called with the segment lock held.
    803  */
    804 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do {			\
    805 	if ((FS)->lfs_version > 1) {					\
    806 		LFS_CLEANERINFO((CIP), (FS), (BP));			\
    807 		(FS)->lfs_freehd = (CIP)->free_head;			\
    808 		brelse(BP, 0);						\
    809 	}								\
    810 	*(FREEP) = (FS)->lfs_freehd;					\
    811 } while (0)
    812 
    813 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do {				\
    814 	(FS)->lfs_freehd = (VAL);					\
    815 	if ((FS)->lfs_version > 1) {					\
    816 		LFS_CLEANERINFO((CIP), (FS), (BP));			\
    817 		(CIP)->free_head = (VAL);				\
    818 		LFS_BWRITE_LOG(BP);					\
    819 		mutex_enter(&lfs_lock);					\
    820 		(FS)->lfs_flags |= LFS_IFDIRTY;				\
    821 		mutex_exit(&lfs_lock);					\
    822 	}								\
    823 } while (0)
    824 
    825 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do {			\
    826 	LFS_CLEANERINFO((CIP), (FS), (BP));				\
    827 	*(FREEP) = (CIP)->free_tail;					\
    828 	brelse(BP, 0);							\
    829 } while (0)
    830 
    831 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do {				\
    832 	LFS_CLEANERINFO((CIP), (FS), (BP));				\
    833 	(CIP)->free_tail = (VAL);					\
    834 	LFS_BWRITE_LOG(BP);						\
    835 	mutex_enter(&lfs_lock);						\
    836 	(FS)->lfs_flags |= LFS_IFDIRTY;					\
    837 	mutex_exit(&lfs_lock);						\
    838 } while (0)
    839 
    840 /*
    841  * On-disk segment summary information
    842  */
    843 typedef struct segsum_v1 SEGSUM_V1;
    844 struct segsum_v1 {
    845 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
    846 	u_int32_t ss_datasum;		/* 4: check sum of data */
    847 	u_int32_t ss_magic;		/* 8: segment summary magic number */
    848 #define SS_MAGIC	0x061561
    849 	int32_t	  ss_next;		/* 12: next segment */
    850 	u_int32_t ss_create;		/* 16: creation time stamp */
    851 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
    852 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
    853 
    854 #define	SS_DIROP	0x01		/* segment begins a dirop */
    855 #define	SS_CONT		0x02		/* more partials to finish this write*/
    856 #define	SS_CLEAN	0x04		/* written by the cleaner */
    857 #define	SS_RFW		0x08		/* written by the roll-forward agent */
    858 #define	SS_RECLAIM	0x10		/* written by the roll-forward agent */
    859 	u_int16_t ss_flags;		/* 24: used for directory operations */
    860 	u_int16_t ss_pad;		/* 26: extra space */
    861 	/* FINFO's and inode daddr's... */
    862 };
    863 
    864 typedef struct segsum SEGSUM;
    865 struct segsum {
    866 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
    867 	u_int32_t ss_datasum;		/* 4: check sum of data */
    868 	u_int32_t ss_magic;		/* 8: segment summary magic number */
    869 	int32_t	  ss_next;		/* 12: next segment */
    870 	u_int32_t ss_ident;		/* 16: roll-forward fsid */
    871 #define ss_ocreate ss_ident /* ident is where create was in v1 */
    872 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
    873 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
    874 	u_int16_t ss_flags;		/* 24: used for directory operations */
    875 	u_int8_t  ss_pad[2];		/* 26: extra space */
    876 	u_int32_t ss_reclino;           /* 28: inode being reclaimed */
    877 	u_int64_t ss_serial;		/* 32: serial number */
    878 	u_int64_t ss_create;		/* 40: time stamp */
    879 	/* FINFO's and inode daddr's... */
    880 };
    881 
    882 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM))
    883 
    884 
    885 /*
    886  * On-disk super block.
    887  */
    888 struct dlfs {
    889 #define	       LFS_MAGIC       0x070162
    890 	u_int32_t dlfs_magic;	  /* 0: magic number */
    891 #define	       LFS_VERSION     2
    892 	u_int32_t dlfs_version;	  /* 4: version number */
    893 
    894 	u_int32_t dlfs_size;	  /* 8: number of blocks in fs (v1) */
    895 				  /*	number of frags in fs (v2) */
    896 	u_int32_t dlfs_ssize;	  /* 12: number of blocks per segment (v1) */
    897 				  /*	 number of bytes per segment (v2) */
    898 	u_int32_t dlfs_dsize;	  /* 16: number of disk blocks in fs */
    899 	u_int32_t dlfs_bsize;	  /* 20: file system block size */
    900 	u_int32_t dlfs_fsize;	  /* 24: size of frag blocks in fs */
    901 	u_int32_t dlfs_frag;	  /* 28: number of frags in a block in fs */
    902 
    903 /* Checkpoint region. */
    904 	u_int32_t dlfs_freehd;	  /* 32: start of the free list */
    905 	int32_t   dlfs_bfree;	  /* 36: number of free disk blocks */
    906 	u_int32_t dlfs_nfiles;	  /* 40: number of allocated inodes */
    907 	int32_t	  dlfs_avail;	  /* 44: blocks available for writing */
    908 	int32_t	  dlfs_uinodes;	  /* 48: inodes in cache not yet on disk */
    909 	int32_t	  dlfs_idaddr;	  /* 52: inode file disk address */
    910 	u_int32_t dlfs_ifile;	  /* 56: inode file inode number */
    911 	int32_t	  dlfs_lastseg;	  /* 60: address of last segment written */
    912 	int32_t	  dlfs_nextseg;	  /* 64: address of next segment to write */
    913 	int32_t	  dlfs_curseg;	  /* 68: current segment being written */
    914 	int32_t	  dlfs_offset;	  /* 72: offset in curseg for next partial */
    915 	int32_t	  dlfs_lastpseg;  /* 76: address of last partial written */
    916 	u_int32_t dlfs_inopf;	  /* 80: v1: time stamp; v2: inodes per frag */
    917 #define dlfs_otstamp dlfs_inopf
    918 
    919 /* These are configuration parameters. */
    920 	u_int32_t dlfs_minfree;	  /* 84: minimum percentage of free blocks */
    921 
    922 /* These fields can be computed from the others. */
    923 	u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */
    924 	u_int32_t dlfs_fsbpseg;	    /* 96: fsb per segment */
    925 	u_int32_t dlfs_inopb;	  /* 100: inodes per block */
    926 	u_int32_t dlfs_ifpb;	  /* 104: IFILE entries per block */
    927 	u_int32_t dlfs_sepb;	  /* 108: SEGUSE entries per block */
    928 	u_int32_t dlfs_nindir;	  /* 112: indirect pointers per block */
    929 	u_int32_t dlfs_nseg;	  /* 116: number of segments */
    930 	u_int32_t dlfs_nspf;	  /* 120: number of sectors per fragment */
    931 	u_int32_t dlfs_cleansz;	  /* 124: cleaner info size in blocks */
    932 	u_int32_t dlfs_segtabsz;  /* 128: segment table size in blocks */
    933 	u_int32_t dlfs_segmask;	  /* 132: calculate offset within a segment */
    934 	u_int32_t dlfs_segshift;  /* 136: fast mult/div for segments */
    935 	u_int32_t dlfs_bshift;	  /* 140: calc block number from file offset */
    936 	u_int32_t dlfs_ffshift;	  /* 144: fast mult/div for frag from file */
    937 	u_int32_t dlfs_fbshift;	  /* 148: fast mult/div for frag from block */
    938 	u_int64_t dlfs_bmask;	  /* 152: calc block offset from file offset */
    939 	u_int64_t dlfs_ffmask;	  /* 160: calc frag offset from file offset */
    940 	u_int64_t dlfs_fbmask;	  /* 168: calc frag offset from block offset */
    941 	u_int32_t dlfs_blktodb;	  /* 176: blktodb and dbtoblk shift constant */
    942 	u_int32_t dlfs_sushift;	  /* 180: fast mult/div for segusage table */
    943 
    944 	int32_t	  dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
    945 #define LFS_MIN_SBINTERVAL     5  /* minimum superblock segment spacing */
    946 #define LFS_MAXNUMSB	       10 /* 188: superblock disk offsets */
    947 	int32_t	   dlfs_sboffs[LFS_MAXNUMSB];
    948 
    949 	u_int32_t dlfs_nclean;	  /* 228: Number of clean segments */
    950 	u_char	  dlfs_fsmnt[MNAMELEN];	 /* 232: name mounted on */
    951 #define LFS_PF_CLEAN 0x1
    952 	u_int16_t dlfs_pflags;	  /* 322: file system persistent flags */
    953 	int32_t	  dlfs_dmeta;	  /* 324: total number of dirty summaries */
    954 	u_int32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */
    955 	u_int32_t dlfs_sumsize;	  /* 332: size of summary blocks */
    956 	u_int64_t dlfs_serial;	  /* 336: serial number */
    957 	u_int32_t dlfs_ibsize;	  /* 344: size of inode blocks */
    958 	int32_t	  dlfs_start;	  /* 348: start of segment 0 */
    959 	u_int64_t dlfs_tstamp;	  /* 352: time stamp */
    960 #define LFS_44INODEFMT 0
    961 #define LFS_MAXINODEFMT 0
    962 	u_int32_t dlfs_inodefmt;  /* 360: inode format version */
    963 	u_int32_t dlfs_interleave; /* 364: segment interleave */
    964 	u_int32_t dlfs_ident;	  /* 368: per-fs identifier */
    965 	u_int32_t dlfs_fsbtodb;	  /* 372: fsbtodb abd dbtodsb shift constant */
    966 	u_int32_t dlfs_resvseg;   /* 376: segments reserved for the cleaner */
    967 	int8_t	  dlfs_pad[128];  /* 380: round to 512 bytes */
    968 /* Checksum -- last valid disk field. */
    969 	u_int32_t dlfs_cksum;	  /* 508: checksum for superblock checking */
    970 };
    971 
    972 /* Type used for the inode bitmap */
    973 typedef u_int32_t lfs_bm_t;
    974 
    975 /*
    976  * Linked list of segments whose byte count needs updating following a
    977  * file truncation.
    978  */
    979 struct segdelta {
    980 	long segnum;
    981 	size_t num;
    982 	LIST_ENTRY(segdelta) list;
    983 };
    984 
    985 /*
    986  * In-memory super block.
    987  */
    988 struct lfs {
    989 	struct dlfs lfs_dlfs;		/* on-disk parameters */
    990 #define lfs_magic lfs_dlfs.dlfs_magic
    991 #define lfs_version lfs_dlfs.dlfs_version
    992 #define lfs_size lfs_dlfs.dlfs_size
    993 #define lfs_ssize lfs_dlfs.dlfs_ssize
    994 #define lfs_dsize lfs_dlfs.dlfs_dsize
    995 #define lfs_bsize lfs_dlfs.dlfs_bsize
    996 #define lfs_fsize lfs_dlfs.dlfs_fsize
    997 #define lfs_frag lfs_dlfs.dlfs_frag
    998 #define lfs_freehd lfs_dlfs.dlfs_freehd
    999 #define lfs_bfree lfs_dlfs.dlfs_bfree
   1000 #define lfs_nfiles lfs_dlfs.dlfs_nfiles
   1001 #define lfs_avail lfs_dlfs.dlfs_avail
   1002 #define lfs_uinodes lfs_dlfs.dlfs_uinodes
   1003 #define lfs_idaddr lfs_dlfs.dlfs_idaddr
   1004 #define lfs_ifile lfs_dlfs.dlfs_ifile
   1005 #define lfs_lastseg lfs_dlfs.dlfs_lastseg
   1006 #define lfs_nextseg lfs_dlfs.dlfs_nextseg
   1007 #define lfs_curseg lfs_dlfs.dlfs_curseg
   1008 #define lfs_offset lfs_dlfs.dlfs_offset
   1009 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg
   1010 #define lfs_otstamp lfs_dlfs.dlfs_inopf
   1011 #define lfs_inopf lfs_dlfs.dlfs_inopf
   1012 #define lfs_minfree lfs_dlfs.dlfs_minfree
   1013 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize
   1014 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg
   1015 #define lfs_inopb lfs_dlfs.dlfs_inopb
   1016 #define lfs_ifpb lfs_dlfs.dlfs_ifpb
   1017 #define lfs_sepb lfs_dlfs.dlfs_sepb
   1018 #define lfs_nindir lfs_dlfs.dlfs_nindir
   1019 #define lfs_nseg lfs_dlfs.dlfs_nseg
   1020 #define lfs_nspf lfs_dlfs.dlfs_nspf
   1021 #define lfs_cleansz lfs_dlfs.dlfs_cleansz
   1022 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz
   1023 #define lfs_segmask lfs_dlfs.dlfs_segmask
   1024 #define lfs_segshift lfs_dlfs.dlfs_segshift
   1025 #define lfs_bmask lfs_dlfs.dlfs_bmask
   1026 #define lfs_bshift lfs_dlfs.dlfs_bshift
   1027 #define lfs_ffmask lfs_dlfs.dlfs_ffmask
   1028 #define lfs_ffshift lfs_dlfs.dlfs_ffshift
   1029 #define lfs_fbmask lfs_dlfs.dlfs_fbmask
   1030 #define lfs_fbshift lfs_dlfs.dlfs_fbshift
   1031 #define lfs_blktodb lfs_dlfs.dlfs_blktodb
   1032 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb
   1033 #define lfs_sushift lfs_dlfs.dlfs_sushift
   1034 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen
   1035 #define lfs_sboffs lfs_dlfs.dlfs_sboffs
   1036 #define lfs_cksum lfs_dlfs.dlfs_cksum
   1037 #define lfs_pflags lfs_dlfs.dlfs_pflags
   1038 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt
   1039 #define lfs_nclean lfs_dlfs.dlfs_nclean
   1040 #define lfs_dmeta lfs_dlfs.dlfs_dmeta
   1041 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg
   1042 #define lfs_sumsize lfs_dlfs.dlfs_sumsize
   1043 #define lfs_serial lfs_dlfs.dlfs_serial
   1044 #define lfs_ibsize lfs_dlfs.dlfs_ibsize
   1045 #define lfs_start lfs_dlfs.dlfs_start
   1046 #define lfs_tstamp lfs_dlfs.dlfs_tstamp
   1047 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt
   1048 #define lfs_interleave lfs_dlfs.dlfs_interleave
   1049 #define lfs_ident lfs_dlfs.dlfs_ident
   1050 #define lfs_resvseg lfs_dlfs.dlfs_resvseg
   1051 
   1052 /* These fields are set at mount time and are meaningless on disk. */
   1053 	struct segment *lfs_sp;		/* current segment being written */
   1054 	struct vnode *lfs_ivnode;	/* vnode for the ifile */
   1055 	u_int32_t  lfs_seglock;		/* single-thread the segment writer */
   1056 	pid_t	  lfs_lockpid;		/* pid of lock holder */
   1057 	lwpid_t	  lfs_locklwp;		/* lwp of lock holder */
   1058 	u_int32_t lfs_iocount;		/* number of ios pending */
   1059 	u_int32_t lfs_writer;		/* don't allow any dirops to start */
   1060 	u_int32_t lfs_dirops;		/* count of active directory ops */
   1061 	u_int32_t lfs_dirvcount;	/* count of VDIROP nodes in this fs */
   1062 	u_int32_t lfs_doifile;		/* Write ifile blocks on next write */
   1063 	u_int32_t lfs_nactive;		/* Number of segments since last ckp */
   1064 	int8_t	  lfs_fmod;		/* super block modified flag */
   1065 	int8_t	  lfs_ronly;		/* mounted read-only flag */
   1066 #define LFS_NOTYET  0x01
   1067 #define LFS_IFDIRTY 0x02
   1068 #define LFS_WARNED  0x04
   1069 #define LFS_UNDIROP 0x08
   1070 	int8_t	  lfs_flags;		/* currently unused flag */
   1071 	u_int16_t lfs_activesb;		/* toggle between superblocks */
   1072 	daddr_t	  lfs_sbactive;		/* disk address of current sb write */
   1073 	struct vnode *lfs_flushvp;	/* vnode being flushed */
   1074 	int lfs_flushvp_fakevref;	/* fake vref count for flushvp */
   1075 	struct vnode *lfs_unlockvp;	/* being inactivated in lfs_segunlock */
   1076 	u_int32_t lfs_diropwait;	/* # procs waiting on dirop flush */
   1077 	size_t lfs_devbsize;		/* Device block size */
   1078 	size_t lfs_devbshift;		/* Device block shift */
   1079 	krwlock_t lfs_fraglock;
   1080 	krwlock_t lfs_iflock;		/* Ifile lock */
   1081 	kcondvar_t lfs_stopcv;		/* Wrap lock */
   1082 	struct lwp *lfs_stoplwp;
   1083 	pid_t lfs_rfpid;		/* Process ID of roll-forward agent */
   1084 	int	  lfs_nadirop;		/* number of active dirop nodes */
   1085 	long	  lfs_ravail;		/* blocks pre-reserved for writing */
   1086 	long	  lfs_favail;		/* blocks pre-reserved for writing */
   1087 	res_t *lfs_resblk;		/* Reserved memory for pageout */
   1088 	TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */
   1089 	TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */
   1090 #define LFS_RESHASH_WIDTH 17
   1091 	LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH];
   1092 	int	  lfs_pdflush;		 /* pagedaemon wants us to flush */
   1093 	u_int32_t **lfs_suflags;	/* Segment use flags */
   1094 #ifdef _KERNEL
   1095 	struct pool lfs_clpool;		/* Pool for struct lfs_cluster */
   1096 	struct pool lfs_bpppool;	/* Pool for bpp */
   1097 	struct pool lfs_segpool;	/* Pool for struct segment */
   1098 #endif /* _KERNEL */
   1099 #define LFS_MAX_CLEANIND 64
   1100 	int32_t  lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */
   1101 	int 	 lfs_cleanind;		/* Index into intervals */
   1102 	int lfs_sleepers;		/* # procs sleeping this fs */
   1103 	int lfs_pages;			/* dirty pages blaming this fs */
   1104 	lfs_bm_t *lfs_ino_bitmap;	/* Inuse inodes bitmap */
   1105 	int lfs_nowrap;			/* Suspend log wrap */
   1106 	int lfs_wrappass;		/* Allow first log wrap requester to pass */
   1107 	int lfs_wrapstatus;		/* Wrap status */
   1108 	int lfs_reclino;		/* Inode being reclaimed */
   1109 	int lfs_startseg;               /* Segment we started writing at */
   1110 	LIST_HEAD(, segdelta) lfs_segdhd;	/* List of pending trunc accounting events */
   1111 };
   1112 
   1113 /* NINDIR is the number of indirects in a file system block. */
   1114 #define	NINDIR(fs)	((fs)->lfs_nindir)
   1115 
   1116 /* INOPB is the number of inodes in a secondary storage block. */
   1117 #define	INOPB(fs)	((fs)->lfs_inopb)
   1118 /* INOPF is the number of inodes in a fragment. */
   1119 #define INOPF(fs)	((fs)->lfs_inopf)
   1120 
   1121 #define	blksize(fs, ip, lbn) \
   1122 	(((lbn) >= ULFS_NDADDR || (ip)->i_ffs1_size >= ((lbn) + 1) << (fs)->lfs_bshift) \
   1123 	    ? (fs)->lfs_bsize \
   1124 	    : (fragroundup(fs, blkoff(fs, (ip)->i_ffs1_size))))
   1125 #define	blkoff(fs, loc)		((int)((loc) & (fs)->lfs_bmask))
   1126 #define fragoff(fs, loc)    /* calculates (loc % fs->lfs_fsize) */ \
   1127     ((int)((loc) & (fs)->lfs_ffmask))
   1128 
   1129 #if defined (_KERNEL)
   1130 #define	fsbtodb(fs, b)		((b) << ((fs)->lfs_ffshift - DEV_BSHIFT))
   1131 #define	dbtofsb(fs, b)		((b) >> ((fs)->lfs_ffshift - DEV_BSHIFT))
   1132 #else
   1133 #define	fsbtodb(fs, b)		((b) << (fs)->lfs_fsbtodb)
   1134 #define	dbtofsb(fs, b)		((b) >> (fs)->lfs_fsbtodb)
   1135 #endif
   1136 
   1137 #define	lblkno(fs, loc)		((loc) >> (fs)->lfs_bshift)
   1138 #define	lblktosize(fs, blk)	((blk) << (fs)->lfs_bshift)
   1139 
   1140 #define fsbtob(fs, b)		((b) << (fs)->lfs_ffshift)
   1141 #define btofsb(fs, b)		((b) >> (fs)->lfs_ffshift)
   1142 
   1143 #define numfrags(fs, loc)	/* calculates (loc / fs->lfs_fsize) */	\
   1144 	((loc) >> (fs)->lfs_ffshift)
   1145 #define blkroundup(fs, size)	/* calculates roundup(size, fs->lfs_bsize) */ \
   1146 	((off_t)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask)))
   1147 #define fragroundup(fs, size)	/* calculates roundup(size, fs->lfs_fsize) */ \
   1148 	((off_t)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask)))
   1149 #define fragstoblks(fs, frags)/* calculates (frags / fs->fs_frag) */ \
   1150 	((frags) >> (fs)->lfs_fbshift)
   1151 #define blkstofrags(fs, blks)	/* calculates (blks * fs->fs_frag) */ \
   1152 	((blks) << (fs)->lfs_fbshift)
   1153 #define fragnum(fs, fsb)	/* calculates (fsb % fs->lfs_frag) */	\
   1154 	((fsb) & ((fs)->lfs_frag - 1))
   1155 #define blknum(fs, fsb)		/* calculates rounddown(fsb, fs->lfs_frag) */ \
   1156 	((fsb) &~ ((fs)->lfs_frag - 1))
   1157 #define dblksize(fs, dp, lbn) \
   1158 	(((lbn) >= ULFS_NDADDR || (dp)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\
   1159 	    ? (fs)->lfs_bsize \
   1160 	    : (fragroundup(fs, blkoff(fs, (dp)->di_size))))
   1161 
   1162 #define	segsize(fs)	((fs)->lfs_version == 1 ?	     		\
   1163 			   lblktosize((fs), (fs)->lfs_ssize) :		\
   1164 			   (fs)->lfs_ssize)
   1165 #define segtod(fs, seg) (((fs)->lfs_version == 1     ?	     		\
   1166 			   (fs)->lfs_ssize << (fs)->lfs_blktodb :	\
   1167 			   btofsb((fs), (fs)->lfs_ssize)) * (seg))
   1168 #define	dtosn(fs, daddr)	/* block address to segment number */	\
   1169 	((uint32_t)(((daddr) - (fs)->lfs_start) / segtod((fs), 1)))
   1170 #define sntod(fs, sn)		/* segment number to disk address */	\
   1171 	((daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start))
   1172 
   1173 /*
   1174  * Structures used by lfs_bmapv and lfs_markv to communicate information
   1175  * about inodes and data blocks.
   1176  */
   1177 typedef struct block_info {
   1178 	u_int32_t bi_inode;		/* inode # */
   1179 	int32_t	bi_lbn;			/* logical block w/in file */
   1180 	int32_t	bi_daddr;		/* disk address of block */
   1181 	u_int64_t bi_segcreate;		/* origin segment create time */
   1182 	int	bi_version;		/* file version number */
   1183 	void	*bi_bp;			/* data buffer */
   1184 	int	bi_size;		/* size of the block (if fragment) */
   1185 } BLOCK_INFO;
   1186 
   1187 /* Compatibility for 1.5 binaries */
   1188 typedef struct block_info_15 {
   1189 	u_int32_t bi_inode;		/* inode # */
   1190 	int32_t	bi_lbn;			/* logical block w/in file */
   1191 	int32_t	bi_daddr;		/* disk address of block */
   1192 	u_int32_t bi_segcreate;		/* origin segment create time */
   1193 	int	bi_version;		/* file version number */
   1194 	void	*bi_bp;			/* data buffer */
   1195 	int	bi_size;		/* size of the block (if fragment) */
   1196 } BLOCK_INFO_15;
   1197 
   1198 /* In-memory description of a segment about to be written. */
   1199 struct segment {
   1200 	struct lfs	 *fs;		/* file system pointer */
   1201 	struct buf	**bpp;		/* pointer to buffer array */
   1202 	struct buf	**cbpp;		/* pointer to next available bp */
   1203 	struct buf	**start_bpp;	/* pointer to first bp in this set */
   1204 	struct buf	 *ibp;		/* buffer pointer to inode page */
   1205 	struct ulfs1_dinode    *idp;          /* pointer to ifile dinode */
   1206 	struct finfo	 *fip;		/* current fileinfo pointer */
   1207 	struct vnode	 *vp;		/* vnode being gathered */
   1208 	void	 *segsum;		/* segment summary info */
   1209 	u_int32_t ninodes;		/* number of inodes in this segment */
   1210 	int32_t seg_bytes_left;		/* bytes left in segment */
   1211 	int32_t sum_bytes_left;		/* bytes left in summary block */
   1212 	u_int32_t seg_number;		/* number of this segment */
   1213 	int32_t *start_lbp;		/* beginning lbn for this set */
   1214 
   1215 #define SEGM_CKP	0x0001		/* doing a checkpoint */
   1216 #define SEGM_CLEAN	0x0002		/* cleaner call; don't sort */
   1217 #define SEGM_SYNC	0x0004		/* wait for segment */
   1218 #define SEGM_PROT	0x0008		/* don't inactivate at segunlock */
   1219 #define SEGM_PAGEDAEMON	0x0010		/* pagedaemon called us */
   1220 #define SEGM_WRITERD	0x0020		/* LFS writed called us */
   1221 #define SEGM_FORCE_CKP	0x0040		/* Force checkpoint right away */
   1222 #define SEGM_RECLAIM	0x0080		/* Writing to reclaim vnode */
   1223 #define SEGM_SINGLE	0x0100		/* Opportunistic writevnodes */
   1224 	u_int16_t seg_flags;		/* run-time flags for this segment */
   1225 	u_int32_t seg_iocount;		/* number of ios pending */
   1226 	int	  ndupino;		/* number of duplicate inodes */
   1227 };
   1228 
   1229 #ifdef _KERNEL
   1230 struct lfs_cluster {
   1231 	size_t bufsize;	       /* Size of kept data */
   1232 	struct buf **bpp;      /* Array of kept buffers */
   1233 	int bufcount;	       /* Number of kept buffers */
   1234 #define LFS_CL_MALLOC	0x00000001
   1235 #define LFS_CL_SHIFT	0x00000002
   1236 #define LFS_CL_SYNC	0x00000004
   1237 	u_int32_t flags;       /* Flags */
   1238 	struct lfs *fs;	       /* LFS that this belongs to */
   1239 	struct segment *seg;   /* Segment structure, for LFS_CL_SYNC */
   1240 };
   1241 
   1242 /*
   1243  * Splay tree containing block numbers allocated through lfs_balloc.
   1244  */
   1245 struct lbnentry {
   1246 	SPLAY_ENTRY(lbnentry) entry;
   1247 	daddr_t lbn;
   1248 };
   1249 #endif /* _KERNEL */
   1250 
   1251 /*
   1252  * LFS inode extensions.
   1253  */
   1254 struct lfs_inode_ext {
   1255 	off_t	  lfs_osize;		/* size of file on disk */
   1256 	u_int32_t lfs_effnblocks;  /* number of blocks when i/o completes */
   1257 	size_t	  lfs_fragsize[ULFS_NDADDR]; /* size of on-disk direct blocks */
   1258 	TAILQ_ENTRY(inode) lfs_dchain;  /* Dirop chain. */
   1259 	TAILQ_ENTRY(inode) lfs_pchain;  /* Paging chain. */
   1260 #define LFSI_NO_GOP_WRITE 0x01
   1261 #define LFSI_DELETED      0x02
   1262 #define LFSI_WRAPBLOCK    0x04
   1263 #define LFSI_WRAPWAIT     0x08
   1264 #define LFSI_BMAP         0x10
   1265 	u_int32_t lfs_iflags;           /* Inode flags */
   1266 	daddr_t   lfs_hiblk;		/* Highest lbn held by inode */
   1267 #ifdef _KERNEL
   1268 	SPLAY_HEAD(lfs_splay, lbnentry) lfs_lbtree; /* Tree of balloc'd lbns */
   1269 	int	  lfs_nbtree;		/* Size of tree */
   1270 	LIST_HEAD(, segdelta) lfs_segdhd;
   1271 #endif
   1272 	int16_t	  lfs_odnlink;		/* on-disk nlink count for cleaner */
   1273 };
   1274 #define i_lfs_osize		inode_ext.lfs->lfs_osize
   1275 #define i_lfs_effnblks		inode_ext.lfs->lfs_effnblocks
   1276 #define i_lfs_fragsize		inode_ext.lfs->lfs_fragsize
   1277 #define i_lfs_dchain		inode_ext.lfs->lfs_dchain
   1278 #define i_lfs_pchain		inode_ext.lfs->lfs_pchain
   1279 #define i_lfs_iflags		inode_ext.lfs->lfs_iflags
   1280 #define i_lfs_hiblk		inode_ext.lfs->lfs_hiblk
   1281 #define i_lfs_lbtree		inode_ext.lfs->lfs_lbtree
   1282 #define i_lfs_nbtree		inode_ext.lfs->lfs_nbtree
   1283 #define i_lfs_segdhd		inode_ext.lfs->lfs_segdhd
   1284 #define i_lfs_odnlink		inode_ext.lfs->lfs_odnlink
   1285 
   1286 /*
   1287  * Macros for determining free space on the disk, with the variable metadata
   1288  * of segment summaries and inode blocks taken into account.
   1289  */
   1290 /*
   1291  * Estimate number of clean blocks not available for writing because
   1292  * they will contain metadata or overhead.  This is calculated as
   1293  *
   1294  *		E = ((C * M / D) * D + (0) * (T - D)) / T
   1295  * or more simply
   1296  *		E = (C * M) / T
   1297  *
   1298  * where
   1299  * C is the clean space,
   1300  * D is the dirty space,
   1301  * M is the dirty metadata, and
   1302  * T = C + D is the total space on disk.
   1303  *
   1304  * This approximates the old formula of E = C * M / D when D is close to T,
   1305  * but avoids falsely reporting "disk full" when the sample size (D) is small.
   1306  */
   1307 #define LFS_EST_CMETA(F) (int32_t)((					\
   1308 	((F)->lfs_dmeta * (int64_t)(F)->lfs_nclean) / 			\
   1309 	((F)->lfs_nseg)))
   1310 
   1311 /* Estimate total size of the disk not including metadata */
   1312 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F))
   1313 
   1314 /* Estimate number of blocks actually available for writing */
   1315 #define LFS_EST_BFREE(F) ((F)->lfs_bfree > LFS_EST_CMETA(F) ?		     \
   1316 			  (F)->lfs_bfree - LFS_EST_CMETA(F) : 0)
   1317 
   1318 /* Amount of non-meta space not available to mortal man */
   1319 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) *			     \
   1320 				   (u_int64_t)(F)->lfs_minfree) /	     \
   1321 				  100)
   1322 
   1323 /* Can credential C write BB blocks */
   1324 #define ISSPACE(F, BB, C)						\
   1325 	((((C) == NOCRED || kauth_cred_geteuid(C) == 0) &&		\
   1326 	  LFS_EST_BFREE(F) >= (BB)) ||					\
   1327 	 (kauth_cred_geteuid(C) != 0 && IS_FREESPACE(F, BB)))
   1328 
   1329 /* Can an ordinary user write BB blocks */
   1330 #define IS_FREESPACE(F, BB)						\
   1331 	  (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F))
   1332 
   1333 /*
   1334  * The minimum number of blocks to create a new inode.  This is:
   1335  * directory direct block (1) + ULFS_NIADDR indirect blocks + inode block (1) +
   1336  * ifile direct block (1) + ULFS_NIADDR indirect blocks = 3 + 2 * ULFS_NIADDR blocks.
   1337  */
   1338 #define LFS_NRESERVE(F) (btofsb((F), (2 * ULFS_NIADDR + 3) << (F)->lfs_bshift))
   1339 
   1340 /* Statistics Counters */
   1341 struct lfs_stats {	/* Must match sysctl list in lfs_vfsops.h ! */
   1342 	u_int	segsused;
   1343 	u_int	psegwrites;
   1344 	u_int	psyncwrites;
   1345 	u_int	pcleanwrites;
   1346 	u_int	blocktot;
   1347 	u_int	cleanblocks;
   1348 	u_int	ncheckpoints;
   1349 	u_int	nwrites;
   1350 	u_int	nsync_writes;
   1351 	u_int	wait_exceeded;
   1352 	u_int	write_exceeded;
   1353 	u_int	flush_invoked;
   1354 	u_int	vflush_invoked;
   1355 	u_int	clean_inlocked;
   1356 	u_int	clean_vnlocked;
   1357 	u_int   segs_reclaimed;
   1358 };
   1359 #ifdef _KERNEL
   1360 extern struct lfs_stats lfs_stats;
   1361 #endif
   1362 
   1363 /* Fcntls to take the place of the lfs syscalls */
   1364 struct lfs_fcntl_markv {
   1365 	BLOCK_INFO *blkiov;	/* blocks to relocate */
   1366 	int blkcnt;		/* number of blocks */
   1367 };
   1368 
   1369 #define LFCNSEGWAITALL	_FCNR_FSPRIV('L', 14, struct timeval)
   1370 #define LFCNSEGWAIT	_FCNR_FSPRIV('L', 15, struct timeval)
   1371 #define LFCNBMAPV	_FCNRW_FSPRIV('L', 2, struct lfs_fcntl_markv)
   1372 #define LFCNMARKV	_FCNRW_FSPRIV('L', 3, struct lfs_fcntl_markv)
   1373 #define LFCNRECLAIM	 _FCNO_FSPRIV('L', 4)
   1374 
   1375 struct lfs_fhandle {
   1376 	char space[28];	/* FHANDLE_SIZE_COMPAT (but used from userland too) */
   1377 };
   1378 #define LFCNREWIND       _FCNR_FSPRIV('L', 6, int)
   1379 #define LFCNINVAL        _FCNR_FSPRIV('L', 7, int)
   1380 #define LFCNRESIZE       _FCNR_FSPRIV('L', 8, int)
   1381 #define LFCNWRAPSTOP	 _FCNR_FSPRIV('L', 9, int)
   1382 #define LFCNWRAPGO	 _FCNR_FSPRIV('L', 10, int)
   1383 #define LFCNIFILEFH	 _FCNW_FSPRIV('L', 11, struct lfs_fhandle)
   1384 #define LFCNWRAPPASS	 _FCNR_FSPRIV('L', 12, int)
   1385 # define LFS_WRAP_GOING   0x0
   1386 # define LFS_WRAP_WAITING 0x1
   1387 #define LFCNWRAPSTATUS	 _FCNW_FSPRIV('L', 13, int)
   1388 
   1389 /*
   1390  * Compat.  Defined for kernel only.  Userland always uses
   1391  * "the one true version".
   1392  */
   1393 #ifdef _KERNEL
   1394 #include <compat/sys/time_types.h>
   1395 
   1396 #define LFCNSEGWAITALL_COMPAT	 _FCNW_FSPRIV('L', 0, struct timeval50)
   1397 #define LFCNSEGWAIT_COMPAT	 _FCNW_FSPRIV('L', 1, struct timeval50)
   1398 #define LFCNIFILEFH_COMPAT	 _FCNW_FSPRIV('L', 5, struct lfs_fhandle)
   1399 #define LFCNIFILEFH_COMPAT2	 _FCN_FSPRIV(F_FSOUT, 'L', 11, 32)
   1400 #define LFCNWRAPSTOP_COMPAT	 _FCNO_FSPRIV('L', 9)
   1401 #define LFCNWRAPGO_COMPAT	 _FCNO_FSPRIV('L', 10)
   1402 #define LFCNSEGWAITALL_COMPAT_50 _FCNR_FSPRIV('L', 0, struct timeval50)
   1403 #define LFCNSEGWAIT_COMPAT_50	 _FCNR_FSPRIV('L', 1, struct timeval50)
   1404 #endif
   1405 
   1406 #ifdef _KERNEL
   1407 /* XXX MP */
   1408 #define	LFS_SEGLOCK_HELD(fs) \
   1409 	((fs)->lfs_seglock != 0 &&					\
   1410 	 (fs)->lfs_lockpid == curproc->p_pid &&				\
   1411 	 (fs)->lfs_locklwp == curlwp->l_lid)
   1412 #endif /* _KERNEL */
   1413 
   1414 /* Debug segment lock */
   1415 #ifdef notyet
   1416 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs))
   1417 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs))
   1418 # define ASSERT_DUNNO_SEGLOCK(fs)
   1419 # define ASSERT_MAYBE_SEGLOCK(fs)
   1420 #else /* !notyet */
   1421 # define ASSERT_DUNNO_SEGLOCK(fs) \
   1422 	DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \
   1423 		LFS_SEGLOCK_HELD(fs)))
   1424 # define ASSERT_SEGLOCK(fs) do {					\
   1425 	if (!LFS_SEGLOCK_HELD(fs)) {					\
   1426 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \
   1427 	}								\
   1428 } while(0)
   1429 # define ASSERT_NO_SEGLOCK(fs) do {					\
   1430 	if (LFS_SEGLOCK_HELD(fs)) {					\
   1431 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \
   1432 	}								\
   1433 } while(0)
   1434 # define ASSERT_MAYBE_SEGLOCK(x)
   1435 #endif /* !notyet */
   1436 
   1437 /*
   1438  * Arguments to mount LFS filesystems
   1439  */
   1440 struct ulfs_args {
   1441 	char	*fspec;			/* block special device to mount */
   1442 };
   1443 
   1444 __BEGIN_DECLS
   1445 void lfs_itimes(struct inode *, const struct timespec *,
   1446     const struct timespec *, const struct timespec *);
   1447 __END_DECLS
   1448 
   1449 #endif /* !_UFS_LFS_LFS_H_ */
   1450