xref: /netbsd-src/sys/ufs/lfs/lfs.h (revision 06be8101a16cc95f40783b3cb7afd12112103a9a)
1 /*	$NetBSD: lfs.h,v 1.37 2001/07/13 20:30:22 perseant Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000 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@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  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *      This product includes software developed by the NetBSD
21  *      Foundation, Inc. and its contributors.
22  * 4. Neither the name of The NetBSD Foundation nor the names of its
23  *    contributors may be used to endorse or promote products derived
24  *    from this software without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 /*-
39  * Copyright (c) 1991, 1993
40  *	The Regents of the University of California.  All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by the University of
53  *	California, Berkeley and its contributors.
54  * 4. Neither the name of the University nor the names of its contributors
55  *    may be used to endorse or promote products derived from this software
56  *    without specific prior written permission.
57  *
58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68  * SUCH DAMAGE.
69  *
70  *	@(#)lfs.h	8.9 (Berkeley) 5/8/95
71  */
72 
73 /*
74  * Compile-time options for LFS.
75  */
76 #define LFS_EAGAIN_FAIL          /* markv fail with EAGAIN if ino is locked */
77 #define LFS_TRACK_IOS            /* attempt to avoid cleaning segments not yet fully written to disk */
78 #define LFS_DEBUG_RFW            /* print roll-forward debugging info */
79 
80 /* #define DEBUG_LFS */              /* Intensive debugging of LFS subsystem */
81 
82 /*
83  * Parameters and generic definitions
84  */
85 #define BW_CLEAN	1
86 #define MIN_FREE_SEGS	2
87 #define LFS_MAX_ACTIVE	10
88 #define LFS_MAXDIROP	(desiredvnodes>>2)
89 #ifndef LFS_ATIME_IFILE
90 # define LFS_ATIME_IFILE 0
91 #endif
92 
93 /*
94  * #define WRITE_THRESHHOLD    ((nbuf >> 1) - 10)
95  * #define WAIT_THRESHHOLD     (nbuf - (nbuf >> 2) - 10)
96  */
97 #define LFS_MAX_BUFS        ((nbuf >> 2) - 10)
98 #define LFS_WAIT_BUFS       ((nbuf >> 1) - (nbuf >> 3) - 10)
99 /* These are new ... is LFS taking up too much memory in its buffers? */
100 #define LFS_MAX_BYTES       (((bufpages >> 2) - 10) * NBPG)
101 #define LFS_WAIT_BYTES      (((bufpages >> 1) - (bufpages >> 3) - 10) * NBPG)
102 #define LFS_BUFWAIT         2
103 
104 #define LFS_LOCK_BUF(bp) do {						\
105 	if (((bp)->b_flags & (B_LOCKED | B_CALL)) == 0) {		\
106 		++locked_queue_count;       				\
107 		locked_queue_bytes += bp->b_bufsize;			\
108 	}								\
109 	(bp)->b_flags |= B_LOCKED;					\
110 } while(0)
111 
112 #define LFS_UNLOCK_BUF(bp) do {						\
113 	if (((bp)->b_flags & (B_LOCKED | B_CALL)) == B_LOCKED) {	\
114 		--locked_queue_count;       				\
115 		locked_queue_bytes -= bp->b_bufsize;			\
116 		if (locked_queue_count < LFS_WAIT_BUFS &&		\
117 		    locked_queue_bytes < LFS_WAIT_BYTES)		\
118 			wakeup(&locked_queue_count);			\
119 	}								\
120 	(bp)->b_flags &= ~B_LOCKED;					\
121 } while(0)
122 
123 /* For convenience */
124 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_ACCESSED|IN_CLEANING)
125 
126 #define LFS_SET_UINO(ip, flags) do {                                    \
127         if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED))   \
128                 ++(ip)->i_lfs->lfs_uinodes;                             \
129         if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING))   \
130                 ++(ip)->i_lfs->lfs_uinodes;                             \
131         if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED))   \
132                 ++(ip)->i_lfs->lfs_uinodes;                             \
133         (ip)->i_flag |= (flags);                                        \
134 } while(0)
135 
136 #define LFS_CLR_UINO(ip, flags) do {                                    \
137         if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED))    \
138                 --(ip)->i_lfs->lfs_uinodes;                             \
139         if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING))    \
140                 --(ip)->i_lfs->lfs_uinodes;                             \
141         if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED))    \
142                 --(ip)->i_lfs->lfs_uinodes;                             \
143         (ip)->i_flag &= ~(flags);                                       \
144 	if ((ip)->i_lfs->lfs_uinodes < 0) {                             \
145 		panic("lfs_uinodes < 0");                               \
146 	}                                                               \
147 } while(0)
148 
149 #define LFS_ITIMES(ip, acc, mod, cre)  do {				\
150        	if ((ip)->i_flag & IN_ACCESS) {                        		\
151 		(ip)->i_ffs_atime = (acc)->tv_sec;			\
152 		(ip)->i_ffs_atimensec = (acc)->tv_nsec;			\
153 		if ((ip)->i_lfs->lfs_version > 1) {			\
154 			struct buf *ibp;				\
155 			IFILE *ifp;					\
156 									\
157 			LFS_IENTRY(ifp, ip->i_lfs, ip->i_number, ibp);	\
158 			ifp->if_atime_sec = (acc)->tv_sec;		\
159 			ifp->if_atime_nsec = (acc)->tv_nsec;		\
160 			VOP_BWRITE(ibp);				\
161 		} else {						\
162 			LFS_SET_UINO(ip, IN_ACCESSED);			\
163 		}                                              		\
164 	}								\
165 	if ((ip)->i_flag & (IN_CHANGE | IN_UPDATE)) {			\
166 		if ((ip)->i_flag & IN_UPDATE) {				\
167 			(ip)->i_ffs_mtime = (mod)->tv_sec;		\
168 			(ip)->i_ffs_mtimensec = (mod)->tv_nsec;		\
169 			(ip)->i_modrev++;				\
170 		}							\
171 		if ((ip)->i_flag & IN_CHANGE) {				\
172 			(ip)->i_ffs_ctime = (cre)->tv_sec;		\
173 			(ip)->i_ffs_ctimensec = (cre)->tv_nsec;		\
174 		}							\
175 		LFS_SET_UINO(ip, IN_MODIFIED);				\
176 	}								\
177 	(ip)->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);		\
178 } while(0)
179 
180 #define WRITEINPROG(vp) (vp->v_dirtyblkhd.lh_first && !(VTOI(vp)->i_flag & \
181 				(IN_MODIFIED | IN_ACCESSED | IN_CLEANING)))
182 
183 /* Here begins the berkeley code */
184 
185 #define	LFS_LABELPAD	8192		/* LFS label size */
186 #define	LFS_SBPAD	8192		/* LFS superblock size */
187 
188 /* On-disk and in-memory checkpoint segment usage structure. */
189 typedef struct segusage SEGUSE;
190 struct segusage {
191 	u_int32_t su_nbytes;		/* 0: number of live bytes */
192 	u_int32_t su_olastmod;		/* 4: SEGUSE last modified timestamp */
193 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
194 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
195 
196 #define	SEGUSE_ACTIVE		0x01	/*  segment currently being written */
197 #define	SEGUSE_DIRTY		0x02	/*  segment has data in it */
198 #define	SEGUSE_SUPERBLOCK	0x04	/*  segment contains a superblock */
199 #define SEGUSE_ERROR            0x08    /*  cleaner: do not clean segment */
200 	u_int32_t su_flags;		/* 12: segment flags */
201 	u_int64_t su_lastmod;		/* 16: last modified timestamp */
202 };
203 
204 typedef struct segusage_v1 SEGUSE_V1;
205 struct segusage_v1 {
206 	u_int32_t su_nbytes;		/* 0: number of live bytes */
207 	u_int32_t su_lastmod;		/* 4: SEGUSE last modified timestamp */
208 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
209 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
210 	u_int32_t su_flags;             /* 12: segment flags  */
211 };
212 
213 #define	SEGUPB(fs)	(fs->lfs_sepb)
214 #define	SEGTABSIZE_SU(fs)						\
215 	(((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb)
216 
217 /* On-disk file information.  One per file with data blocks in the segment. */
218 typedef struct finfo FINFO;
219 struct finfo {
220 	u_int32_t fi_nblocks;		/* number of blocks */
221 	u_int32_t fi_version;		/* version number */
222 	u_int32_t fi_ino;		/* inode number */
223 	u_int32_t fi_lastlength;	/* length of last block in array */
224 	ufs_daddr_t fi_blocks[1];	/* array of logical block numbers */
225 };
226 
227 
228 /* On-disk super block. */
229 struct dlfs {
230 #define        LFS_MAGIC       0x070162
231         u_int32_t dlfs_magic;     /* 0: magic number */
232 #define        LFS_VERSION     2
233         u_int32_t dlfs_version;   /* 4: version number */
234 
235         u_int32_t dlfs_size;      /* 8: number of blocks in fs (v1) */
236 				  /*    number of frags in fs (v2) */
237         u_int32_t dlfs_ssize;     /* 12: number of blocks per segment (v1) */
238 	                          /*     number of bytes per segment (v2) */
239         u_int32_t dlfs_dsize;     /* 16: number of disk blocks in fs */
240         u_int32_t dlfs_bsize;     /* 20: file system block size */
241         u_int32_t dlfs_fsize;     /* 24: size of frag blocks in fs */
242         u_int32_t dlfs_frag;      /* 28: number of frags in a block in fs */
243 
244 /* Checkpoint region. */
245         u_int32_t dlfs_free;      /* 32: start of the free list */
246         u_int32_t dlfs_bfree;     /* 36: number of free disk blocks */
247         u_int32_t dlfs_nfiles;    /* 40: number of allocated inodes */
248         int32_t   dlfs_avail;     /* 44: blocks available for writing */
249         int32_t   dlfs_uinodes;   /* 48: inodes in cache not yet on disk */
250         ufs_daddr_t  dlfs_idaddr; /* 52: inode file disk address */
251         u_int32_t dlfs_ifile;     /* 56: inode file inode number */
252         ufs_daddr_t  dlfs_lastseg; /* 60: address of last segment written */
253         ufs_daddr_t  dlfs_nextseg; /* 64: address of next segment to write */
254         ufs_daddr_t  dlfs_curseg; /* 68: current segment being written */
255         ufs_daddr_t  dlfs_offset; /* 72: offset in curseg for next partial */
256         ufs_daddr_t  dlfs_lastpseg; /* 76: address of last partial written */
257 	u_int32_t dlfs_inopf;     /* 80: v1: time stamp; v2: inodes per frag */
258 #define dlfs_otstamp dlfs_inopf
259 
260 /* These are configuration parameters. */
261         u_int32_t dlfs_minfree;   /* 84: minimum percentage of free blocks */
262 
263 /* These fields can be computed from the others. */
264         u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */
265         u_int32_t dlfs_fsbpseg;     /* 96: fsb per segment */
266         u_int32_t dlfs_inopb;     /* 100: inodes per block */
267         u_int32_t dlfs_ifpb;      /* 104: IFILE entries per block */
268         u_int32_t dlfs_sepb;      /* 108: SEGUSE entries per block */
269         u_int32_t dlfs_nindir;    /* 112: indirect pointers per block */
270         u_int32_t dlfs_nseg;      /* 116: number of segments */
271         u_int32_t dlfs_nspf;      /* 120: number of sectors per fragment */
272         u_int32_t dlfs_cleansz;   /* 124: cleaner info size in blocks */
273         u_int32_t dlfs_segtabsz;  /* 128: segment table size in blocks */
274         u_int32_t dlfs_segmask;   /* 132: calculate offset within a segment */
275         u_int32_t dlfs_segshift;  /* 136: fast mult/div for segments */
276         u_int32_t dlfs_bshift;    /* 140: calc block number from file offset */
277         u_int32_t dlfs_ffshift;   /* 144: fast mult/div for frag from file */
278         u_int32_t dlfs_fbshift;   /* 148: fast mult/div for frag from block */
279         u_int64_t dlfs_bmask;     /* 152: calc block offset from file offset */
280         u_int64_t dlfs_ffmask;    /* 160: calc frag offset from file offset */
281         u_int64_t dlfs_fbmask;    /* 168: calc frag offset from block offset */
282         u_int32_t dlfs_blktodb;   /* 176: blktodb and dbtoblk shift constant */
283         u_int32_t dlfs_sushift;   /* 180: fast mult/div for segusage table */
284 
285         int32_t   dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
286 #define LFS_MIN_SBINTERVAL      5  /* minimum superblock segment spacing */
287 #define LFS_MAXNUMSB            10 /* 188: superblock disk offsets */
288         ufs_daddr_t       dlfs_sboffs[LFS_MAXNUMSB];
289 
290 	u_int32_t dlfs_nclean;    /* 228: Number of clean segments */
291 	u_char	  dlfs_fsmnt[MNAMELEN];	 /* 232: name mounted on */
292 #define LFS_PF_CLEAN 0x1
293 	u_int16_t dlfs_pflags;    /* 322: file system persistent flags */
294 	int32_t   dlfs_dmeta;     /* 324: total number of dirty summaries */
295 	u_int32_t dlfs_minfreeseg; /* 328: segs reserved for cleaner */
296 	u_int32_t dlfs_sumsize;   /* 332: size of summary blocks */
297 	u_int64_t dlfs_serial;    /* 336: serial number */
298 	u_int32_t dlfs_ibsize;    /* 344: size of inode blocks */
299 	ufs_daddr_t dlfs_start;   /* 348: start of segment 0 */
300 	u_int64_t dlfs_tstamp;    /* 352: time stamp */
301 #define LFS_44INODEFMT 0
302 #define LFS_MAXINODEFMT 0
303 	u_int32_t dlfs_inodefmt;  /* 360: inode format version */
304 	u_int32_t dlfs_interleave; /* 364: segment interleave */
305 	u_int32_t dlfs_ident;     /* 368: per-fs identifier */
306 	u_int32_t dlfs_fsbtodb;   /* 372: fsbtodb abd dbtodsb shift constant */
307 	int8_t    dlfs_pad[132];  /* 376: round to 512 bytes */
308 /* Checksum -- last valid disk field. */
309 	u_int32_t dlfs_cksum;     /* 508: checksum for superblock checking */
310 };
311 
312 /* Maximum number of io's we can have pending at once */
313 #define LFS_THROTTLE  16 /* XXX should be better paramtrized - ? */
314 
315 /* In-memory super block. */
316 struct lfs {
317         struct dlfs lfs_dlfs;           /* on-disk parameters */
318 #define lfs_magic lfs_dlfs.dlfs_magic
319 #define lfs_version lfs_dlfs.dlfs_version
320 #define lfs_size lfs_dlfs.dlfs_size
321 #define lfs_ssize lfs_dlfs.dlfs_ssize
322 #define lfs_dsize lfs_dlfs.dlfs_dsize
323 #define lfs_bsize lfs_dlfs.dlfs_bsize
324 #define lfs_fsize lfs_dlfs.dlfs_fsize
325 #define lfs_frag lfs_dlfs.dlfs_frag
326 #define lfs_free lfs_dlfs.dlfs_free
327 #define lfs_bfree lfs_dlfs.dlfs_bfree
328 #define lfs_nfiles lfs_dlfs.dlfs_nfiles
329 #define lfs_avail lfs_dlfs.dlfs_avail
330 #define lfs_uinodes lfs_dlfs.dlfs_uinodes
331 #define lfs_idaddr lfs_dlfs.dlfs_idaddr
332 #define lfs_ifile lfs_dlfs.dlfs_ifile
333 #define lfs_lastseg lfs_dlfs.dlfs_lastseg
334 #define lfs_nextseg lfs_dlfs.dlfs_nextseg
335 #define lfs_curseg lfs_dlfs.dlfs_curseg
336 #define lfs_offset lfs_dlfs.dlfs_offset
337 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg
338 #define lfs_otstamp lfs_dlfs.dlfs_inopf
339 #define lfs_inopf lfs_dlfs.dlfs_inopf
340 #define lfs_minfree lfs_dlfs.dlfs_minfree
341 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize
342 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg
343 #define lfs_inopb lfs_dlfs.dlfs_inopb
344 #define lfs_ifpb lfs_dlfs.dlfs_ifpb
345 #define lfs_sepb lfs_dlfs.dlfs_sepb
346 #define lfs_nindir lfs_dlfs.dlfs_nindir
347 #define lfs_nseg lfs_dlfs.dlfs_nseg
348 #define lfs_nspf lfs_dlfs.dlfs_nspf
349 #define lfs_cleansz lfs_dlfs.dlfs_cleansz
350 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz
351 #define lfs_segmask lfs_dlfs.dlfs_segmask
352 #define lfs_segshift lfs_dlfs.dlfs_segshift
353 #define lfs_bmask lfs_dlfs.dlfs_bmask
354 #define lfs_bshift lfs_dlfs.dlfs_bshift
355 #define lfs_ffmask lfs_dlfs.dlfs_ffmask
356 #define lfs_ffshift lfs_dlfs.dlfs_ffshift
357 #define lfs_fbmask lfs_dlfs.dlfs_fbmask
358 #define lfs_fbshift lfs_dlfs.dlfs_fbshift
359 #define lfs_blktodb lfs_dlfs.dlfs_blktodb
360 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb
361 #define lfs_sushift lfs_dlfs.dlfs_sushift
362 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen
363 #define lfs_sboffs lfs_dlfs.dlfs_sboffs
364 #define lfs_cksum lfs_dlfs.dlfs_cksum
365 #define lfs_pflags lfs_dlfs.dlfs_pflags
366 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt
367 #define lfs_nclean lfs_dlfs.dlfs_nclean
368 #define lfs_dmeta lfs_dlfs.dlfs_dmeta
369 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg
370 #define lfs_sumsize lfs_dlfs.dlfs_sumsize
371 #define lfs_serial lfs_dlfs.dlfs_serial
372 #define lfs_ibsize lfs_dlfs.dlfs_ibsize
373 #define lfs_start lfs_dlfs.dlfs_start
374 #define lfs_tstamp lfs_dlfs.dlfs_tstamp
375 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt
376 #define lfs_interleave lfs_dlfs.dlfs_interleave
377 #define lfs_ident lfs_dlfs.dlfs_ident
378 
379 /* These fields are set at mount time and are meaningless on disk. */
380 	struct segment *lfs_sp;		/* current segment being written */
381 	struct vnode *lfs_ivnode;	/* vnode for the ifile */
382 	u_int32_t  lfs_seglock;		/* single-thread the segment writer */
383 	pid_t	  lfs_lockpid;		/* pid of lock holder */
384 	u_int32_t lfs_iocount;		/* number of ios pending */
385 	u_int32_t lfs_writer;		/* don't allow any dirops to start */
386 	u_int32_t lfs_dirops;		/* count of active directory ops */
387 	u_int32_t lfs_doifile;		/* Write ifile blocks on next write */
388 	u_int32_t lfs_nactive;		/* Number of segments since last ckp */
389 	int8_t	  lfs_fmod;		/* super block modified flag */
390 	int8_t	  lfs_ronly;		/* mounted read-only flag */
391 #define LFS_NOTYET 0x01
392 	int8_t	  lfs_flags;		/* currently unused flag */
393 	u_int16_t lfs_activesb;         /* toggle between superblocks */
394 #ifdef LFS_TRACK_IOS
395 	daddr_t   lfs_pending[LFS_THROTTLE]; /* daddrs of pending writes */
396 #endif /* LFS_TRACK_IOS */
397 	daddr_t   lfs_sbactive;         /* disk address of in-progress sb write */
398 	struct vnode *lfs_flushvp;      /* vnode being flushed */
399 	struct vnode *lfs_unlockvp;     /* being inactivated in lfs_segunlock */
400 	u_int32_t lfs_diropwait;	/* # procs waiting on dirop flush */
401 	size_t lfs_devbsize;		/* Device block size */
402 	size_t lfs_devbshift;		/* Device block shift */
403 	struct lock lfs_freelock;
404 	pid_t lfs_rfpid;		/* Process ID of roll-forward agent */
405 	int       lfs_nadirop;		/* number of active dirop nodes */
406 	long      lfs_ravail;           /* blocks pre-reserved for writing */
407 };
408 
409 /*
410  * Inode 0:	out-of-band inode number
411  * Inode 1:	IFILE inode number
412  * Inode 2:	root inode
413  * Inode 3:	lost+found inode number
414  */
415 #define	LFS_UNUSED_INUM	0		/* out of band inode number */
416 #define	LFS_IFILE_INUM	1		/* IFILE inode number */
417 #define	LOSTFOUNDINO	3		/* lost+found inode number */
418 #define	LFS_FIRST_INUM	4		/* first free inode number */
419 
420 /* Address calculations for metadata located in the inode */
421 #define	S_INDIR(fs)	-NDADDR
422 #define	D_INDIR(fs)	(S_INDIR(fs) - NINDIR(fs) - 1)
423 #define	T_INDIR(fs)	(D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1)
424 
425 /* Unassigned disk addresses. */
426 #define	UNASSIGNED	-1
427 #define UNWRITTEN       -2
428 
429 /* Unused logical block number */
430 #define LFS_UNUSED_LBN	-1
431 
432 typedef struct ifile IFILE;
433 struct ifile {
434 	u_int32_t if_version;		/* inode version number */
435 #define	LFS_UNUSED_DADDR	0	/* out-of-band daddr */
436 	ufs_daddr_t if_daddr;		/* inode disk address */
437 	ino_t	  if_nextfree;		/* next-unallocated inode */
438 	/* XXX - when inode format changes, this changes too */
439 	u_int32_t if_atime_sec;		/* Last access time, seconds */
440 	u_int32_t if_atime_nsec;	/* and nanoseconds */
441 };
442 
443 typedef struct ifile_v1 IFILE_V1;
444 struct ifile_v1 {
445 	u_int32_t if_version;		/* inode version number */
446 	ufs_daddr_t if_daddr;		/* inode disk address */
447 	ino_t	  if_nextfree;		/* next-unallocated inode */
448 #if LFS_ATIME_IFILE
449 	struct timespec if_atime;	/* Last access time */
450 #endif
451 };
452 
453 /*
454  * Cleaner information structure.  This resides in the ifile and is used
455  * to pass information between the cleaner and the kernel.
456  */
457 typedef struct _cleanerinfo {
458 	u_int32_t clean;		/* number of clean segments */
459 	u_int32_t dirty;		/* number of dirty segments */
460 	u_int32_t bfree;		/* disk blocks free */
461 	int32_t   avail;		/* disk blocks available */
462 	u_int32_t free_head;            /* head of the inode free list */
463 	u_int32_t free_tail;            /* tail of the inode free list */
464 } CLEANERINFO;
465 
466 #define	CLEANSIZE_SU(fs)						\
467 	((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift)
468 
469 /*
470  * All summary blocks are the same size, so we can always read a summary
471  * block easily from a segment.
472  */
473 #define	LFS_V1_SUMMARY_SIZE	512
474 #define	LFS_DFL_SUMMARY_SIZE	512
475 
476 /* On-disk segment summary information */
477 typedef struct segsum_v1 SEGSUM_V1;
478 struct segsum_v1 {
479 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
480 	u_int32_t ss_datasum;		/* 4: check sum of data */
481 	u_int32_t ss_magic;		/* 8: segment summary magic number */
482 #define SS_MAGIC	0x061561
483 	ufs_daddr_t ss_next;		/* 12: next segment */
484 	u_int32_t ss_create;		/* 16: creation time stamp */
485 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
486 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
487 
488 #define	SS_DIROP	0x01		/* segment begins a dirop */
489 #define	SS_CONT		0x02		/* more partials to finish this write*/
490 	u_int16_t ss_flags;		/* 24: used for directory operations */
491 	u_int16_t ss_pad;		/* 26: extra space */
492 	/* FINFO's and inode daddr's... */
493 };
494 
495 typedef struct segsum SEGSUM;
496 struct segsum {
497 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
498 	u_int32_t ss_datasum;		/* 4: check sum of data */
499 	u_int32_t ss_magic;		/* 8: segment summary magic number */
500 	ufs_daddr_t ss_next;		/* 12: next segment */
501 	u_int32_t ss_ident;		/* 16: roll-forward fsid */
502 #define ss_ocreate ss_ident /* ident is where create was in v1 */
503 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
504 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
505 	u_int16_t ss_flags;		/* 24: used for directory operations */
506 	u_int8_t  ss_pad[6];		/* 26: extra space */
507 	u_int64_t ss_serial;		/* 32: serial number */
508 	u_int64_t ss_create;		/* 40: time stamp */
509 	/* FINFO's and inode daddr's... */
510 };
511 
512 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM))
513 
514 /* NINDIR is the number of indirects in a file system block. */
515 #define	NINDIR(fs)	((fs)->lfs_nindir)
516 
517 /* INOPB is the number of inodes in a secondary storage block. */
518 #define	INOPB(fs)	((fs)->lfs_inopb)
519 /* INOPF is the number of inodes in a fragment. */
520 #define INOPF(fs)       ((fs)->lfs_inopf)
521 
522 #define	blksize(fs, ip, lbn) \
523 	(((lbn) >= NDADDR || (ip)->i_ffs_size >= ((lbn) + 1) << (fs)->lfs_bshift) \
524 	    ? (fs)->lfs_bsize \
525 	    : (fragroundup(fs, blkoff(fs, (ip)->i_ffs_size))))
526 #define	blkoff(fs, loc)		((int)(loc) & (fs)->lfs_bmask)
527 #define fragoff(fs, loc)    /* calculates (loc % fs->lfs_fsize) */ \
528     ((int)((loc) & (fs)->lfs_ffmask))
529 #define	fsbtodb(fs, b)		((b) << (fs)->lfs_fsbtodb)
530 #define	dbtofsb(fs, b)		((b) >> (fs)->lfs_fsbtodb)
531 #define fragstodb(fs, b)	((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift))
532 #define dbtofrags(fs, b)	((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift))
533 #define	lblkno(fs, loc)		((loc) >> (fs)->lfs_bshift)
534 #define	lblktosize(fs, blk)	((blk) << (fs)->lfs_bshift)
535 /* Same as above, but named like dbtob(), btodb() */
536 #define fsbtob(fs, b)		((b) << ((fs)->lfs_bshift - \
537 				(fs)->lfs_blktodb + (fs)->lfs_fsbtodb))
538 #define btofsb(fs, b)		((b) >> ((fs)->lfs_bshift - \
539 				(fs)->lfs_blktodb + (fs)->lfs_fsbtodb))
540 #define fsbtofrags(fs, b)	((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \
541 				(fs)->lfs_fsbtodb))
542 #define fragstofsb(fs, b)	((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \
543 				(fs)->lfs_fsbtodb))
544 #define btofrags(fs, b)		((b) >> (fs)->lfs_ffshift)
545 #define numfrags(fs, loc)	/* calculates (loc / fs->lfs_fsize) */	\
546 	((loc) >> (fs)->lfs_ffshift)
547 #define blkroundup(fs, size)	/* calculates roundup(size, fs->lfs_bsize) */ \
548 	((int)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask)))
549 #define fragroundup(fs, size)	/* calculates roundup(size, fs->lfs_fsize) */ \
550 	((int)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask)))
551 #define fragstoblks(fs, frags)	/* calculates (frags / fs->lfs_frag) */ \
552 	((frags) >> (fs)->lfs_fbshift)
553 #define blkstofrags(fs, blks)	/* calculates (blks * fs->lfs_frag) */ \
554 	((blks) << (fs)->lfs_fbshift)
555 #define fragnum(fs, fsb)	/* calculates (fsb % fs->lfs_frag) */ \
556 	((fsb) & ((fs)->lfs_frag - 1))
557 #define blknum(fs, fsb)		/* calculates rounddown(fsb, fs->lfs_frag) */ \
558 	((fsb) &~ ((fs)->lfs_frag - 1))
559 #define dblksize(fs, dip, lbn) \
560 	(((lbn) >= NDADDR || (dip)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\
561 	    ? (fs)->lfs_bsize \
562 	    : (fragroundup(fs, blkoff(fs, (dip)->di_size))))
563 
564 #define segtod(fs, seg) (((fs)->lfs_version == 1     ?       \
565 			   (fs)->lfs_ssize << (fs)->lfs_blktodb :       \
566 			   btofsb((fs), (fs)->lfs_ssize)) * (seg))
567 #define	dtosn(fs, daddr)	/* block address to segment number */	\
568 	(((daddr) - (fs)->lfs_start) / segtod((fs), 1))
569 #define sntod(fs, sn) 		/* segment number to disk address */	\
570 	((ufs_daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start))
571 
572 /* Read in the block with the cleaner info from the ifile. */
573 #define LFS_CLEANERINFO(CP, F, BP) {					\
574 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
575 	if (bread((F)->lfs_ivnode,					\
576 	    (ufs_daddr_t)0, (F)->lfs_bsize, NOCRED, &(BP)))		\
577 		panic("lfs: ifile read");				\
578 	(CP) = (CLEANERINFO *)(BP)->b_data;				\
579 }
580 
581 /* Synchronize the Ifile cleaner info with current avail and bfree */
582 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do {                \
583     if ((w) || (cip)->bfree != (fs)->lfs_bfree ||                \
584         (cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail) {    \
585 	(cip)->bfree = (fs)->lfs_bfree;                          \
586         (cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail;       \
587 	(void) VOP_BWRITE(bp); /* Ifile */                       \
588     } else                                                       \
589 	brelse(bp);                                              \
590 } while(0)
591 
592 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do {                       \
593 	if ((FS)->lfs_version > 1) {                                    \
594 		LFS_CLEANERINFO((CIP), (FS), (BP));                     \
595 		(FS)->lfs_free = (CIP)->free_head;			\
596 		brelse(BP);                                             \
597 	}								\
598 	*(FREEP) = (FS)->lfs_free;					\
599 } while (0)
600 
601 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do {                         \
602 	(FS)->lfs_free = (VAL);						\
603 	if ((FS)->lfs_version > 1) {                                    \
604 		LFS_CLEANERINFO((CIP), (FS), (BP));                     \
605 		(CIP)->free_head = (VAL);                 		\
606 		VOP_BWRITE(BP);                                         \
607 	}                                                               \
608 } while (0)
609 
610 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do {                       \
611 	LFS_CLEANERINFO((CIP), (FS), (BP));                     	\
612 	*(FREEP) = (CIP)->free_tail;					\
613 	brelse(BP);                                             	\
614 } while (0)
615 
616 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do {                         \
617 	LFS_CLEANERINFO((CIP), (FS), (BP));                     	\
618 	(CIP)->free_tail = (VAL);                 			\
619 	VOP_BWRITE(BP);                                         	\
620 } while (0)
621 
622 /*
623  * XXX - v1 compatibility code is not allowed to touch if_atime, since it
624  * may not be mapped!
625  */
626 /* Read in the block with a specific inode from the ifile. */
627 #define	LFS_IENTRY(IP, F, IN, BP) {					\
628 	int _e;								\
629 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
630 	if ((_e = bread((F)->lfs_ivnode,				\
631     	(IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz,	\
632     	(F)->lfs_bsize, NOCRED, &(BP))) != 0)				\
633 		panic("lfs: ifile read %d", _e);			\
634 	if((F)->lfs_version == 1)					\
635 		(IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data + (IN) % (F)->lfs_ifpb); \
636 	else								\
637 		(IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb;	\
638 }
639 
640 /* Read in the block with a specific segment usage entry from the ifile. */
641 #define	LFS_SEGENTRY(SP, F, IN, BP) {					\
642 	int _e;								\
643 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
644 	if ((_e = bread((F)->lfs_ivnode,				\
645 	    ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz,			\
646 	    (F)->lfs_bsize, NOCRED, &(BP))) != 0)			\
647 		panic("lfs: ifile read: %d", _e);			\
648 	if ((F)->lfs_version == 1)					\
649 		(SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data +		\
650 			((IN) & ((F)->lfs_sepb - 1)));			\
651 	else								\
652 		(SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb);	\
653 }
654 
655 /* Determine if a buffer belongs to the ifile */
656 #define IS_IFILE(bp)	(VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM)
657 
658 /*
659  * Structures used by lfs_bmapv and lfs_markv to communicate information
660  * about inodes and data blocks.
661  */
662 typedef struct block_info {
663 	ino_t	bi_inode;		/* inode # */
664 	ufs_daddr_t bi_lbn;		/* logical block w/in file */
665 	ufs_daddr_t bi_daddr;		/* disk address of block */
666 	u_int64_t   bi_segcreate;	/* origin segment create time */
667 	int	bi_version;		/* file version number */
668 	void	*bi_bp;			/* data buffer */
669 	int     bi_size;		/* size of the block (if fragment) */
670 } BLOCK_INFO;
671 
672 /* Compatibility for 1.5 binaries */
673 typedef struct block_info_15 {
674 	ino_t	bi_inode;		/* inode # */
675 	ufs_daddr_t bi_lbn;		/* logical block w/in file */
676 	ufs_daddr_t bi_daddr;		/* disk address of block */
677 	u_int32_t   bi_segcreate;	/* origin segment create time */
678 	int	bi_version;		/* file version number */
679 	void	*bi_bp;			/* data buffer */
680 	int     bi_size;		/* size of the block (if fragment) */
681 } BLOCK_INFO_15;
682 
683 /* In-memory description of a segment about to be written. */
684 struct segment {
685 	struct lfs	 *fs;		/* file system pointer */
686 	struct buf	**bpp;		/* pointer to buffer array */
687 	struct buf	**cbpp;		/* pointer to next available bp */
688 	struct buf	**start_bpp;	/* pointer to first bp in this set */
689 	struct buf	 *ibp;		/* buffer pointer to inode page */
690 	struct dinode    *idp;          /* pointer to ifile dinode */
691 	struct finfo	 *fip;		/* current fileinfo pointer */
692 	struct vnode	 *vp;		/* vnode being gathered */
693 	void	 *segsum;		/* segment summary info */
694 	u_int32_t ninodes;		/* number of inodes in this segment */
695 	u_int32_t seg_bytes_left;	/* bytes left in segment */
696 	u_int32_t sum_bytes_left;	/* bytes left in summary block */
697 	u_int32_t seg_number;		/* number of this segment */
698 	ufs_daddr_t *start_lbp;		/* beginning lbn for this set */
699 
700 #define	SEGM_CKP	0x01		/* doing a checkpoint */
701 #define	SEGM_CLEAN	0x02		/* cleaner call; don't sort */
702 #define	SEGM_SYNC	0x04		/* wait for segment */
703 #define	SEGM_PROT	0x08		/* don't inactivate at segunlock */
704 	u_int16_t seg_flags;		/* run-time flags for this segment */
705 };
706 
707 /*
708  * Macros for determining free space on the disk, with the variable metadata
709  * of segment summaries and inode blocks taken into account.
710  */
711 /* Estimate number of clean blocks not available for writing */
712 #define LFS_EST_CMETA(F) (int32_t)((((F)->lfs_dmeta *                        \
713 				     (int64_t)(F)->lfs_nclean) /             \
714 				      ((F)->lfs_nseg - (F)->lfs_nclean)))
715 
716 /* Estimate total size of the disk not including metadata */
717 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F))
718 
719 /* Estimate number of blocks actually available for writing */
720 #define LFS_EST_BFREE(F) ((F)->lfs_bfree - LFS_EST_CMETA(F) - (F)->lfs_dmeta)
721 
722 /* Amount of non-meta space not available to mortal man */
723 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) *                      \
724                                    (u_int64_t)(F)->lfs_minfree) /            \
725 			          100)
726 
727 /* Can credential C write BB blocks */
728 #define ISSPACE(F, BB, C)						\
729 	((((C) == NOCRED || (C)->cr_uid == 0) &&			\
730           LFS_EST_BFREE(F) >= (BB)) ||					\
731 	 ((C)->cr_uid != 0 && IS_FREESPACE(F, BB)))
732 
733 /* Can an ordinary user write BB blocks */
734 #define IS_FREESPACE(F, BB)						\
735           (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F))
736 
737 /* Statistics Counters */
738 struct lfs_stats {
739 	u_int	segsused;
740 	u_int	psegwrites;
741 	u_int	psyncwrites;
742 	u_int	pcleanwrites;
743 	u_int	blocktot;
744 	u_int	cleanblocks;
745 	u_int	ncheckpoints;
746 	u_int	nwrites;
747 	u_int	nsync_writes;
748 	u_int	wait_exceeded;
749 	u_int	write_exceeded;
750 	u_int	flush_invoked;
751 	u_int	vflush_invoked;
752 };
753 extern struct lfs_stats lfs_stats;
754