xref: /netbsd-src/sys/ufs/lfs/lfs.h (revision 27527e67bbdf8d9ec84fd58803048ed6d181ece2)
1 /*	$NetBSD: lfs.h,v 1.95 2005/12/11 12:25:26 christos 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@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. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)lfs.h	8.9 (Berkeley) 5/8/95
67  */
68 
69 #ifndef _UFS_LFS_LFS_H_
70 #define _UFS_LFS_LFS_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	LOSTFOUNDINO	3		/* 3: lost+found inode number */
89 #define	LFS_FIRST_INUM	4		/* 4: first free inode number */
90 
91 #define	LFS_V1_SUMMARY_SIZE	512     /* V1 fixed summary size */
92 #define	LFS_DFL_SUMMARY_SIZE	512	/* Default summary size */
93 
94 #define LFS_MAX_DADDR	0x7fffffff	/* Highest addressable fsb */
95 
96 #define LFS_MAXNAMLEN	255		/* maximum name length in a dir */
97 
98 /* Adjustable filesystem parameters */
99 #define MIN_FREE_SEGS	2
100 #ifndef LFS_ATIME_IFILE
101 # define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */
102 #endif
103 #define LFS_MARKV_MAXBLKCNT	65536	/* Max block count for lfs_markv() */
104 
105 /* Misc. definitions */
106 #define BW_CLEAN	1		/* Flag for lfs_bwrite_ext() */
107 #define PG_DELWRI	PG_PAGER1	/* Local def for delayed pageout */
108 
109 /* Resource limits */
110 #define LFS_MAX_BUFS	    ((nbuf >> 2) - 10)
111 #define LFS_WAIT_BUFS	    ((nbuf >> 1) - (nbuf >> 3) - 10)
112 #define LFS_INVERSE_MAX_BUFS(n) (((n) + 10) << 2)
113 #define LFS_MAX_BYTES	    ((bufmem_lowater >> 2) - 10 * PAGE_SIZE)
114 #define LFS_INVERSE_MAX_BYTES(n) (((n) + 10 * PAGE_SIZE) << 2)
115 #define LFS_WAIT_BYTES	    ((bufmem_lowater >> 1) - (bufmem_lowater >> 3) - 10 * PAGE_SIZE)
116 #define LFS_MAX_DIROP	    ((desiredvnodes >> 2) + (desiredvnodes >> 3))
117 #define LFS_MAX_PAGES \
118      (((uvmexp.active + uvmexp.inactive + uvmexp.free) * uvmexp.filemin) >> 8)
119 #define LFS_WAIT_PAGES \
120      (((uvmexp.active + uvmexp.inactive + uvmexp.free) * uvmexp.filemax) >> 8)
121 #define LFS_BUFWAIT	    2	/* How long to wait if over *_WAIT_* */
122 
123 /*
124  * Reserved blocks for lfs_malloc
125  */
126 
127 /* Structure to keep reserved blocks */
128 typedef struct lfs_res_blk {
129 	void *p;
130 	LIST_ENTRY(lfs_res_blk) res;
131 	int size;
132 	char inuse;
133 } res_t;
134 
135 /* Types for lfs_newbuf and lfs_malloc */
136 #define LFS_NB_UNKNOWN -1
137 #define LFS_NB_SUMMARY	0
138 #define LFS_NB_SBLOCK	1
139 #define LFS_NB_IBLOCK	2
140 #define LFS_NB_CLUSTER	3
141 #define LFS_NB_CLEAN	4
142 #define LFS_NB_BLKIOV	5
143 #define LFS_NB_COUNT	6 /* always last */
144 
145 /* Number of reserved memory blocks of each type */
146 #define LFS_N_SUMMARIES 2
147 #define LFS_N_SBLOCKS	1   /* Always 1, to throttle superblock writes */
148 #define LFS_N_IBLOCKS	16  /* In theory ssize/bsize; in practice around 2 */
149 #define LFS_N_CLUSTERS	16  /* In theory ssize/MAXPHYS */
150 #define LFS_N_CLEAN	0
151 #define LFS_N_BLKIOV	1
152 
153 /* Total count of "large" (non-pool) types */
154 #define LFS_N_TOTAL (LFS_N_SUMMARIES + LFS_N_SBLOCKS + LFS_N_IBLOCKS +	\
155 		     LFS_N_CLUSTERS + LFS_N_CLEAN + LFS_N_BLKIOV)
156 
157 /* Counts for pool types */
158 #define LFS_N_CL	LFS_N_CLUSTERS
159 #define LFS_N_BPP	2
160 #define LFS_N_SEG	2
161 
162 /*
163  * "struct buf" associated definitions
164  */
165 
166 /* Unassigned disk addresses. */
167 #define	UNASSIGNED	-1
168 #define UNWRITTEN	-2
169 
170 /* Unused logical block number */
171 #define LFS_UNUSED_LBN	-1
172 
173 /* Determine if a buffer belongs to the ifile */
174 #define IS_IFILE(bp)	(VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM)
175 
176 # define LFS_LOCK_BUF(bp) do {						\
177 	if (((bp)->b_flags & (B_LOCKED | B_CALL)) == 0) {		\
178 		simple_lock(&lfs_subsys_lock);				\
179 		++locked_queue_count;					\
180 		locked_queue_bytes += bp->b_bufsize;			\
181 		simple_unlock(&lfs_subsys_lock);			\
182 	}								\
183 	(bp)->b_flags |= B_LOCKED;					\
184 } while (0)
185 
186 # define LFS_UNLOCK_BUF(bp) do {					\
187 	if (((bp)->b_flags & (B_LOCKED | B_CALL)) == B_LOCKED) {	\
188 		simple_lock(&lfs_subsys_lock);				\
189 		--locked_queue_count;					\
190 		locked_queue_bytes -= bp->b_bufsize;			\
191 		if (locked_queue_count < LFS_WAIT_BUFS &&		\
192 		    locked_queue_bytes < LFS_WAIT_BYTES)		\
193 			wakeup(&locked_queue_count);			\
194 		simple_unlock(&lfs_subsys_lock);			\
195 	}								\
196 	(bp)->b_flags &= ~B_LOCKED;					\
197 } while (0)
198 
199 #ifdef _KERNEL
200 
201 extern u_long bufmem_lowater, bufmem_hiwater; /* XXX */
202 
203 # define LFS_IS_MALLOC_BUF(bp) (((bp)->b_flags & B_CALL) &&		\
204      (bp)->b_iodone == lfs_callback)
205 
206 # ifdef DEBUG
207 #  define LFS_DEBUG_COUNTLOCKED(m) do {					\
208 	if (lfs_debug_log_subsys[DLOG_LLIST]) {				\
209 		lfs_countlocked(&locked_queue_count, &locked_queue_bytes, (m)); \
210 		wakeup(&locked_queue_count);				\
211 	}								\
212 } while (0)
213 # else
214 #  define LFS_DEBUG_COUNTLOCKED(m)
215 # endif
216 
217 /* log for debugging writes to the Ifile */
218 # ifdef DEBUG
219 struct lfs_log_entry {
220 	const char *op;
221 	const char *file;
222 	int pid;
223 	int line;
224 	daddr_t block;
225 	unsigned long flags;
226 };
227 extern int lfs_lognum;
228 extern struct lfs_log_entry lfs_log[LFS_LOGLENGTH];
229 #  define LFS_BWRITE_LOG(bp) lfs_bwrite_log((bp), __FILE__, __LINE__)
230 #  define LFS_ENTER_LOG(theop, thefile, theline, lbn, theflags, thepid) do {\
231 	int _s;								\
232 									\
233 	simple_lock(&lfs_subsys_lock);					\
234 	_s = splbio();							\
235 	lfs_log[lfs_lognum].op = theop;					\
236 	lfs_log[lfs_lognum].file = thefile;				\
237 	lfs_log[lfs_lognum].line = (theline);				\
238 	lfs_log[lfs_lognum].pid = (thepid);				\
239 	lfs_log[lfs_lognum].block = (lbn);				\
240 	lfs_log[lfs_lognum].flags = (theflags);				\
241 	lfs_lognum = (lfs_lognum + 1) % LFS_LOGLENGTH;			\
242 	splx(_s);							\
243 	simple_unlock(&lfs_subsys_lock);				\
244 } while (0)
245 
246 #  define LFS_BCLEAN_LOG(fs, bp) do {					\
247 	if ((bp)->b_vp == (fs)->lfs_ivnode)				\
248 		LFS_ENTER_LOG("clear", __FILE__, __LINE__,		\
249 			      bp->b_lblkno, bp->b_flags, curproc->p_pid);\
250 } while (0)
251 
252 /* Must match list in lfs_vfsops.c ! */
253 #  define DLOG_RF     0  /* roll forward */
254 #  define DLOG_ALLOC  1  /* inode alloc */
255 #  define DLOG_AVAIL  2  /* lfs_{,r,f}avail */
256 #  define DLOG_FLUSH  3  /* flush */
257 #  define DLOG_LLIST  4  /* locked list accounting */
258 #  define DLOG_WVNODE 5  /* vflush/writevnodes verbose */
259 #  define DLOG_VNODE  6  /* vflush/writevnodes */
260 #  define DLOG_SEG    7  /* segwrite */
261 #  define DLOG_SU     8  /* seguse accounting */
262 #  define DLOG_CLEAN  9  /* cleaner routines */
263 #  define DLOG_MOUNT  10 /* mount/unmount */
264 #  define DLOG_PAGE   11 /* putpages/gop_write */
265 #  define DLOG_DIROP  12 /* dirop accounting */
266 #  define DLOG_MALLOC 13 /* lfs_malloc accounting */
267 #  define DLOG_MAX    14 /* The terminator */
268 #  define DLOG(a) lfs_debug_log a
269 # else /* ! DEBUG */
270 #  define LFS_BCLEAN_LOG(fs, bp)
271 #  define LFS_BWRITE_LOG(bp)		VOP_BWRITE((bp))
272 #  define DLOG(a)
273 # endif /* ! DEBUG */
274 #else /* ! _KERNEL */
275 # define LFS_BWRITE_LOG(bp)		VOP_BWRITE((bp))
276 #endif /* _KERNEL */
277 
278 #ifdef _KERNEL
279 /* Filehandle structure for exported LFSes */
280 struct lfid {
281 	struct ufid lfid_ufid;
282 #define lfid_len lfid_ufid.ufid_len
283 #define lfid_ino lfid_ufid.ufid_ino
284 #define lfid_gen lfid_ufid.ufid_gen
285 	uint32_t lfid_ident;
286 };
287 #endif /* _KERNEL */
288 
289 /*
290  * "struct inode" associated definitions
291  */
292 
293 /* Address calculations for metadata located in the inode */
294 #define	S_INDIR(fs)	-NDADDR
295 #define	D_INDIR(fs)	(S_INDIR(fs) - NINDIR(fs) - 1)
296 #define	T_INDIR(fs)	(D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1)
297 
298 /* For convenience */
299 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING)
300 
301 #define LFS_SET_UINO(ip, flags) do {					\
302 	simple_lock(&(ip)->i_lfs->lfs_interlock);			\
303 	if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED))	\
304 		++(ip)->i_lfs->lfs_uinodes;				\
305 	if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING))	\
306 		++(ip)->i_lfs->lfs_uinodes;				\
307 	if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED))	\
308 		++(ip)->i_lfs->lfs_uinodes;				\
309 	(ip)->i_flag |= (flags);					\
310 	simple_unlock(&(ip)->i_lfs->lfs_interlock);			\
311 } while (0)
312 
313 #define LFS_CLR_UINO(ip, flags) do {					\
314 	simple_lock(&(ip)->i_lfs->lfs_interlock);			\
315 	if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED))	\
316 		--(ip)->i_lfs->lfs_uinodes;				\
317 	if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING))	\
318 		--(ip)->i_lfs->lfs_uinodes;				\
319 	if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED))	\
320 		--(ip)->i_lfs->lfs_uinodes;				\
321 	(ip)->i_flag &= ~(flags);					\
322 	if ((ip)->i_lfs->lfs_uinodes < 0) {				\
323 		panic("lfs_uinodes < 0");				\
324 	}								\
325 	simple_unlock(&(ip)->i_lfs->lfs_interlock);			\
326 } while (0)
327 
328 #define LFS_ITIMES(ip, acc, mod, cre) \
329 	while ((ip)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY)) \
330 		lfs_itimes(ip, acc, mod, cre)
331 
332 /*
333  * "struct vnode" associated definitions
334  */
335 
336 /* Heuristic emptiness measure */
337 #define VPISEMPTY(vp)	 (LIST_EMPTY(&(vp)->v_dirtyblkhd) && 		\
338 			  !((vp)->v_flag & VONWORKLST) &&		\
339 			  VTOI(vp)->i_lfs_nbtree == 0)
340 
341 /* XXX Shouldn't we use v_numoutput instead? */
342 #define WRITEINPROG(vp) (!LIST_EMPTY(&(vp)->v_dirtyblkhd) &&		\
343 		!(VTOI(vp)->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING)))
344 
345 
346 /*
347  * On-disk and in-memory checkpoint segment usage structure.
348  */
349 typedef struct segusage SEGUSE;
350 struct segusage {
351 	u_int32_t su_nbytes;		/* 0: number of live bytes */
352 	u_int32_t su_olastmod;		/* 4: SEGUSE last modified timestamp */
353 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
354 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
355 
356 #define	SEGUSE_ACTIVE		0x01	/*  segment currently being written */
357 #define	SEGUSE_DIRTY		0x02	/*  segment has data in it */
358 #define	SEGUSE_SUPERBLOCK	0x04	/*  segment contains a superblock */
359 #define SEGUSE_ERROR		0x08	/*  cleaner: do not clean segment */
360 #define SEGUSE_EMPTY		0x10	/*  segment is empty */
361 #define SEGUSE_INVAL		0x20	/*  segment is invalid */
362 	u_int32_t su_flags;		/* 12: segment flags */
363 	u_int64_t su_lastmod;		/* 16: last modified timestamp */
364 };
365 
366 typedef struct segusage_v1 SEGUSE_V1;
367 struct segusage_v1 {
368 	u_int32_t su_nbytes;		/* 0: number of live bytes */
369 	u_int32_t su_lastmod;		/* 4: SEGUSE last modified timestamp */
370 	u_int16_t su_nsums;		/* 8: number of summaries in segment */
371 	u_int16_t su_ninos;		/* 10: number of inode blocks in seg */
372 	u_int32_t su_flags;		/* 12: segment flags  */
373 };
374 
375 #define	SEGUPB(fs)	(fs->lfs_sepb)
376 #define	SEGTABSIZE_SU(fs)						\
377 	(((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb)
378 
379 #ifdef _KERNEL
380 # define SHARE_IFLOCK(F) 						\
381   do {									\
382 	simple_lock(&(F)->lfs_interlock);				\
383 	lockmgr(&(F)->lfs_iflock, LK_SHARED, &(F)->lfs_interlock);	\
384 	simple_unlock(&(F)->lfs_interlock);				\
385   } while(0)
386 # define UNSHARE_IFLOCK(F)						\
387   do {									\
388 	simple_lock(&(F)->lfs_interlock);				\
389 	lockmgr(&(F)->lfs_iflock, LK_RELEASE, &(F)->lfs_interlock);	\
390 	simple_unlock(&(F)->lfs_interlock);				\
391   } while(0)
392 #else /* ! _KERNEL */
393 # define SHARE_IFLOCK(F)
394 # define UNSHARE_IFLOCK(F)
395 #endif /* ! _KERNEL */
396 
397 /* Read in the block with a specific segment usage entry from the ifile. */
398 #define	LFS_SEGENTRY(SP, F, IN, BP) do {				\
399 	int _e;								\
400 	SHARE_IFLOCK(F);						\
401 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
402 	if ((_e = bread((F)->lfs_ivnode,				\
403 	    ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz,			\
404 	    (F)->lfs_bsize, NOCRED, &(BP))) != 0)			\
405 		panic("lfs: ifile read: %d", _e);			\
406 	if ((F)->lfs_version == 1)					\
407 		(SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data +		\
408 			((IN) & ((F)->lfs_sepb - 1)));			\
409 	else								\
410 		(SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb);	\
411 	UNSHARE_IFLOCK(F);						\
412 } while (0)
413 
414 #define LFS_WRITESEGENTRY(SP, F, IN, BP) do {				\
415 	if ((SP)->su_nbytes == 0)					\
416 		(SP)->su_flags |= SEGUSE_EMPTY;				\
417 	else								\
418 		(SP)->su_flags &= ~SEGUSE_EMPTY;			\
419 	(F)->lfs_suflags[(F)->lfs_activesb][(IN)] = (SP)->su_flags;	\
420 	LFS_BWRITE_LOG(BP);						\
421 } while (0)
422 
423 /*
424  * On-disk file information.  One per file with data blocks in the segment.
425  */
426 typedef struct finfo FINFO;
427 struct finfo {
428 	u_int32_t fi_nblocks;		/* number of blocks */
429 	u_int32_t fi_version;		/* version number */
430 	u_int32_t fi_ino;		/* inode number */
431 	u_int32_t fi_lastlength;	/* length of last block in array */
432 	int32_t	  fi_blocks[1];		/* array of logical block numbers */
433 };
434 /* sizeof FINFO except fi_blocks */
435 #define	FINFOSIZE	(sizeof(FINFO) - sizeof(int32_t))
436 
437 /*
438  * Index file inode entries.
439  */
440 typedef struct ifile IFILE;
441 struct ifile {
442 	u_int32_t if_version;		/* inode version number */
443 #define	LFS_UNUSED_DADDR	0	/* out-of-band daddr */
444 	int32_t	  if_daddr;		/* inode disk address */
445 	u_int32_t if_nextfree;		/* next-unallocated inode */
446 	u_int32_t if_atime_sec;		/* Last access time, seconds */
447 	u_int32_t if_atime_nsec;	/* and nanoseconds */
448 };
449 
450 typedef struct ifile_v1 IFILE_V1;
451 struct ifile_v1 {
452 	u_int32_t if_version;		/* inode version number */
453 	int32_t	  if_daddr;		/* inode disk address */
454 	u_int32_t if_nextfree;		/* next-unallocated inode */
455 #if LFS_ATIME_IFILE
456 	struct timespec if_atime;	/* Last access time */
457 #endif
458 };
459 
460 /*
461  * LFSv1 compatibility code is not allowed to touch if_atime, since it
462  * may not be mapped!
463  */
464 /* Read in the block with a specific inode from the ifile. */
465 #define	LFS_IENTRY(IP, F, IN, BP) do {					\
466 	int _e;								\
467 	SHARE_IFLOCK(F);						\
468 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
469 	if ((_e = bread((F)->lfs_ivnode,				\
470 	(IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz,	\
471 	(F)->lfs_bsize, NOCRED, &(BP))) != 0)				\
472 		panic("lfs: ifile read %d", _e);			\
473 	if ((F)->lfs_version == 1)					\
474 		(IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data +		\
475 				 (IN) % (F)->lfs_ifpb);			\
476 	else								\
477 		(IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb;	\
478 	UNSHARE_IFLOCK(F);						\
479 } while (0)
480 
481 /*
482  * Cleaner information structure.  This resides in the ifile and is used
483  * to pass information from the kernel to the cleaner.
484  */
485 typedef struct _cleanerinfo {
486 	u_int32_t clean;		/* number of clean segments */
487 	u_int32_t dirty;		/* number of dirty segments */
488 	int32_t   bfree;		/* disk blocks free */
489 	int32_t	  avail;		/* disk blocks available */
490 	u_int32_t free_head;		/* head of the inode free list */
491 	u_int32_t free_tail;		/* tail of the inode free list */
492 } CLEANERINFO;
493 
494 #define	CLEANSIZE_SU(fs)						\
495 	((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift)
496 
497 /* Read in the block with the cleaner info from the ifile. */
498 #define LFS_CLEANERINFO(CP, F, BP) do {					\
499 	SHARE_IFLOCK(F);						\
500 	VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS;			\
501 	if (bread((F)->lfs_ivnode,					\
502 	    (daddr_t)0, (F)->lfs_bsize, NOCRED, &(BP)))			\
503 		panic("lfs: ifile read");				\
504 	(CP) = (CLEANERINFO *)(BP)->b_data;				\
505 	UNSHARE_IFLOCK(F);						\
506 } while (0)
507 
508 /*
509  * Synchronize the Ifile cleaner info with current avail and bfree.
510  */
511 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do {		 	\
512     simple_lock(&(fs)->lfs_interlock);					\
513     if ((w) || (cip)->bfree != (fs)->lfs_bfree ||		 	\
514 	(cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail - 		\
515 	(fs)->lfs_favail) {	 					\
516 	(cip)->bfree = (fs)->lfs_bfree;				 	\
517 	(cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail -		\
518 		(fs)->lfs_favail;				 	\
519 	if (((bp)->b_flags & B_GATHERED) == 0) {		 	\
520 		(fs)->lfs_flags |= LFS_IFDIRTY;			 	\
521 	}								\
522 	simple_unlock(&(fs)->lfs_interlock);				\
523 	(void) LFS_BWRITE_LOG(bp); /* Ifile */			 	\
524     } else {							 	\
525 	simple_unlock(&(fs)->lfs_interlock);				\
526 	brelse(bp);						 	\
527     }									\
528 } while (0)
529 
530 /*
531  * Get the head of the inode free list.
532  * Always caled with the segment lock held.
533  */
534 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do {			\
535 	if ((FS)->lfs_version > 1) {					\
536 		LFS_CLEANERINFO((CIP), (FS), (BP));			\
537 		(FS)->lfs_freehd = (CIP)->free_head;			\
538 		brelse(BP);						\
539 	}								\
540 	*(FREEP) = (FS)->lfs_freehd;					\
541 } while (0)
542 
543 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do {				\
544 	(FS)->lfs_freehd = (VAL);					\
545 	if ((FS)->lfs_version > 1) {					\
546 		LFS_CLEANERINFO((CIP), (FS), (BP));			\
547 		(CIP)->free_head = (VAL);				\
548 		LFS_BWRITE_LOG(BP);					\
549 		simple_lock(&fs->lfs_interlock);			\
550 		(FS)->lfs_flags |= LFS_IFDIRTY;				\
551 		simple_unlock(&fs->lfs_interlock);			\
552 	}								\
553 } while (0)
554 
555 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do {			\
556 	LFS_CLEANERINFO((CIP), (FS), (BP));				\
557 	*(FREEP) = (CIP)->free_tail;					\
558 	brelse(BP);							\
559 } while (0)
560 
561 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do {				\
562 	LFS_CLEANERINFO((CIP), (FS), (BP));				\
563 	(CIP)->free_tail = (VAL);					\
564 	LFS_BWRITE_LOG(BP);						\
565 	simple_lock(&fs->lfs_interlock);				\
566 	(FS)->lfs_flags |= LFS_IFDIRTY;					\
567 	simple_unlock(&fs->lfs_interlock);				\
568 } while (0)
569 
570 /*
571  * On-disk segment summary information
572  */
573 typedef struct segsum_v1 SEGSUM_V1;
574 struct segsum_v1 {
575 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
576 	u_int32_t ss_datasum;		/* 4: check sum of data */
577 	u_int32_t ss_magic;		/* 8: segment summary magic number */
578 #define SS_MAGIC	0x061561
579 	int32_t	  ss_next;		/* 12: next segment */
580 	u_int32_t ss_create;		/* 16: creation time stamp */
581 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
582 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
583 
584 #define	SS_DIROP	0x01		/* segment begins a dirop */
585 #define	SS_CONT		0x02		/* more partials to finish this write*/
586 	u_int16_t ss_flags;		/* 24: used for directory operations */
587 	u_int16_t ss_pad;		/* 26: extra space */
588 	/* FINFO's and inode daddr's... */
589 };
590 
591 typedef struct segsum SEGSUM;
592 struct segsum {
593 	u_int32_t ss_sumsum;		/* 0: check sum of summary block */
594 	u_int32_t ss_datasum;		/* 4: check sum of data */
595 	u_int32_t ss_magic;		/* 8: segment summary magic number */
596 	int32_t	  ss_next;		/* 12: next segment */
597 	u_int32_t ss_ident;		/* 16: roll-forward fsid */
598 #define ss_ocreate ss_ident /* ident is where create was in v1 */
599 	u_int16_t ss_nfinfo;		/* 20: number of file info structures */
600 	u_int16_t ss_ninos;		/* 22: number of inodes in summary */
601 	u_int16_t ss_flags;		/* 24: used for directory operations */
602 	u_int8_t  ss_pad[6];		/* 26: extra space */
603 	u_int64_t ss_serial;		/* 32: serial number */
604 	u_int64_t ss_create;		/* 40: time stamp */
605 	/* FINFO's and inode daddr's... */
606 };
607 
608 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM))
609 
610 
611 /*
612  * On-disk super block.
613  */
614 struct dlfs {
615 #define	       LFS_MAGIC       0x070162
616 	u_int32_t dlfs_magic;	  /* 0: magic number */
617 #define	       LFS_VERSION     2
618 	u_int32_t dlfs_version;	  /* 4: version number */
619 
620 	u_int32_t dlfs_size;	  /* 8: number of blocks in fs (v1) */
621 				  /*	number of frags in fs (v2) */
622 	u_int32_t dlfs_ssize;	  /* 12: number of blocks per segment (v1) */
623 				  /*	 number of bytes per segment (v2) */
624 	u_int32_t dlfs_dsize;	  /* 16: number of disk blocks in fs */
625 	u_int32_t dlfs_bsize;	  /* 20: file system block size */
626 	u_int32_t dlfs_fsize;	  /* 24: size of frag blocks in fs */
627 	u_int32_t dlfs_frag;	  /* 28: number of frags in a block in fs */
628 
629 /* Checkpoint region. */
630 	u_int32_t dlfs_freehd;	  /* 32: start of the free list */
631 	int32_t   dlfs_bfree;	  /* 36: number of free disk blocks */
632 	u_int32_t dlfs_nfiles;	  /* 40: number of allocated inodes */
633 	int32_t	  dlfs_avail;	  /* 44: blocks available for writing */
634 	int32_t	  dlfs_uinodes;	  /* 48: inodes in cache not yet on disk */
635 	int32_t	  dlfs_idaddr;	  /* 52: inode file disk address */
636 	u_int32_t dlfs_ifile;	  /* 56: inode file inode number */
637 	int32_t	  dlfs_lastseg;	  /* 60: address of last segment written */
638 	int32_t	  dlfs_nextseg;	  /* 64: address of next segment to write */
639 	int32_t	  dlfs_curseg;	  /* 68: current segment being written */
640 	int32_t	  dlfs_offset;	  /* 72: offset in curseg for next partial */
641 	int32_t	  dlfs_lastpseg;  /* 76: address of last partial written */
642 	u_int32_t dlfs_inopf;	  /* 80: v1: time stamp; v2: inodes per frag */
643 #define dlfs_otstamp dlfs_inopf
644 
645 /* These are configuration parameters. */
646 	u_int32_t dlfs_minfree;	  /* 84: minimum percentage of free blocks */
647 
648 /* These fields can be computed from the others. */
649 	u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */
650 	u_int32_t dlfs_fsbpseg;	    /* 96: fsb per segment */
651 	u_int32_t dlfs_inopb;	  /* 100: inodes per block */
652 	u_int32_t dlfs_ifpb;	  /* 104: IFILE entries per block */
653 	u_int32_t dlfs_sepb;	  /* 108: SEGUSE entries per block */
654 	u_int32_t dlfs_nindir;	  /* 112: indirect pointers per block */
655 	u_int32_t dlfs_nseg;	  /* 116: number of segments */
656 	u_int32_t dlfs_nspf;	  /* 120: number of sectors per fragment */
657 	u_int32_t dlfs_cleansz;	  /* 124: cleaner info size in blocks */
658 	u_int32_t dlfs_segtabsz;  /* 128: segment table size in blocks */
659 	u_int32_t dlfs_segmask;	  /* 132: calculate offset within a segment */
660 	u_int32_t dlfs_segshift;  /* 136: fast mult/div for segments */
661 	u_int32_t dlfs_bshift;	  /* 140: calc block number from file offset */
662 	u_int32_t dlfs_ffshift;	  /* 144: fast mult/div for frag from file */
663 	u_int32_t dlfs_fbshift;	  /* 148: fast mult/div for frag from block */
664 	u_int64_t dlfs_bmask;	  /* 152: calc block offset from file offset */
665 	u_int64_t dlfs_ffmask;	  /* 160: calc frag offset from file offset */
666 	u_int64_t dlfs_fbmask;	  /* 168: calc frag offset from block offset */
667 	u_int32_t dlfs_blktodb;	  /* 176: blktodb and dbtoblk shift constant */
668 	u_int32_t dlfs_sushift;	  /* 180: fast mult/div for segusage table */
669 
670 	int32_t	  dlfs_maxsymlinklen; /* 184: max length of an internal symlink */
671 #define LFS_MIN_SBINTERVAL     5  /* minimum superblock segment spacing */
672 #define LFS_MAXNUMSB	       10 /* 188: superblock disk offsets */
673 	int32_t	   dlfs_sboffs[LFS_MAXNUMSB];
674 
675 	u_int32_t dlfs_nclean;	  /* 228: Number of clean segments */
676 	u_char	  dlfs_fsmnt[MNAMELEN];	 /* 232: name mounted on */
677 #define LFS_PF_CLEAN 0x1
678 	u_int16_t dlfs_pflags;	  /* 322: file system persistent flags */
679 	int32_t	  dlfs_dmeta;	  /* 324: total number of dirty summaries */
680 	u_int32_t dlfs_minfreeseg; /* 328: segs reserved for cleaner */
681 	u_int32_t dlfs_sumsize;	  /* 332: size of summary blocks */
682 	u_int64_t dlfs_serial;	  /* 336: serial number */
683 	u_int32_t dlfs_ibsize;	  /* 344: size of inode blocks */
684 	int32_t	  dlfs_start;	  /* 348: start of segment 0 */
685 	u_int64_t dlfs_tstamp;	  /* 352: time stamp */
686 #define LFS_44INODEFMT 0
687 #define LFS_MAXINODEFMT 0
688 	u_int32_t dlfs_inodefmt;  /* 360: inode format version */
689 	u_int32_t dlfs_interleave; /* 364: segment interleave */
690 	u_int32_t dlfs_ident;	  /* 368: per-fs identifier */
691 	u_int32_t dlfs_fsbtodb;	  /* 372: fsbtodb abd dbtodsb shift constant */
692 	int8_t	  dlfs_pad[132];  /* 376: round to 512 bytes */
693 /* Checksum -- last valid disk field. */
694 	u_int32_t dlfs_cksum;	  /* 508: checksum for superblock checking */
695 };
696 
697 /*
698  * In-memory super block.
699  */
700 struct lfs {
701 	struct dlfs lfs_dlfs;		/* on-disk parameters */
702 #define lfs_magic lfs_dlfs.dlfs_magic
703 #define lfs_version lfs_dlfs.dlfs_version
704 #define lfs_size lfs_dlfs.dlfs_size
705 #define lfs_ssize lfs_dlfs.dlfs_ssize
706 #define lfs_dsize lfs_dlfs.dlfs_dsize
707 #define lfs_bsize lfs_dlfs.dlfs_bsize
708 #define lfs_fsize lfs_dlfs.dlfs_fsize
709 #define lfs_frag lfs_dlfs.dlfs_frag
710 #define lfs_freehd lfs_dlfs.dlfs_freehd
711 #define lfs_bfree lfs_dlfs.dlfs_bfree
712 #define lfs_nfiles lfs_dlfs.dlfs_nfiles
713 #define lfs_avail lfs_dlfs.dlfs_avail
714 #define lfs_uinodes lfs_dlfs.dlfs_uinodes
715 #define lfs_idaddr lfs_dlfs.dlfs_idaddr
716 #define lfs_ifile lfs_dlfs.dlfs_ifile
717 #define lfs_lastseg lfs_dlfs.dlfs_lastseg
718 #define lfs_nextseg lfs_dlfs.dlfs_nextseg
719 #define lfs_curseg lfs_dlfs.dlfs_curseg
720 #define lfs_offset lfs_dlfs.dlfs_offset
721 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg
722 #define lfs_otstamp lfs_dlfs.dlfs_inopf
723 #define lfs_inopf lfs_dlfs.dlfs_inopf
724 #define lfs_minfree lfs_dlfs.dlfs_minfree
725 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize
726 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg
727 #define lfs_inopb lfs_dlfs.dlfs_inopb
728 #define lfs_ifpb lfs_dlfs.dlfs_ifpb
729 #define lfs_sepb lfs_dlfs.dlfs_sepb
730 #define lfs_nindir lfs_dlfs.dlfs_nindir
731 #define lfs_nseg lfs_dlfs.dlfs_nseg
732 #define lfs_nspf lfs_dlfs.dlfs_nspf
733 #define lfs_cleansz lfs_dlfs.dlfs_cleansz
734 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz
735 #define lfs_segmask lfs_dlfs.dlfs_segmask
736 #define lfs_segshift lfs_dlfs.dlfs_segshift
737 #define lfs_bmask lfs_dlfs.dlfs_bmask
738 #define lfs_bshift lfs_dlfs.dlfs_bshift
739 #define lfs_ffmask lfs_dlfs.dlfs_ffmask
740 #define lfs_ffshift lfs_dlfs.dlfs_ffshift
741 #define lfs_fbmask lfs_dlfs.dlfs_fbmask
742 #define lfs_fbshift lfs_dlfs.dlfs_fbshift
743 #define lfs_blktodb lfs_dlfs.dlfs_blktodb
744 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb
745 #define lfs_sushift lfs_dlfs.dlfs_sushift
746 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen
747 #define lfs_sboffs lfs_dlfs.dlfs_sboffs
748 #define lfs_cksum lfs_dlfs.dlfs_cksum
749 #define lfs_pflags lfs_dlfs.dlfs_pflags
750 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt
751 #define lfs_nclean lfs_dlfs.dlfs_nclean
752 #define lfs_dmeta lfs_dlfs.dlfs_dmeta
753 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg
754 #define lfs_sumsize lfs_dlfs.dlfs_sumsize
755 #define lfs_serial lfs_dlfs.dlfs_serial
756 #define lfs_ibsize lfs_dlfs.dlfs_ibsize
757 #define lfs_start lfs_dlfs.dlfs_start
758 #define lfs_tstamp lfs_dlfs.dlfs_tstamp
759 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt
760 #define lfs_interleave lfs_dlfs.dlfs_interleave
761 #define lfs_ident lfs_dlfs.dlfs_ident
762 
763 /* These fields are set at mount time and are meaningless on disk. */
764 	struct segment *lfs_sp;		/* current segment being written */
765 	struct vnode *lfs_ivnode;	/* vnode for the ifile */
766 	u_int32_t  lfs_seglock;		/* single-thread the segment writer */
767 	pid_t	  lfs_lockpid;		/* pid of lock holder */
768 	u_int32_t lfs_iocount;		/* number of ios pending */
769 	u_int32_t lfs_writer;		/* don't allow any dirops to start */
770 	u_int32_t lfs_dirops;		/* count of active directory ops */
771 	u_int32_t lfs_doifile;		/* Write ifile blocks on next write */
772 	u_int32_t lfs_nactive;		/* Number of segments since last ckp */
773 	int8_t	  lfs_fmod;		/* super block modified flag */
774 	int8_t	  lfs_ronly;		/* mounted read-only flag */
775 #define LFS_NOTYET  0x01
776 #define LFS_IFDIRTY 0x02
777 #define LFS_WARNED  0x04
778 #define LFS_UNDIROP 0x08
779 	int8_t	  lfs_flags;		/* currently unused flag */
780 	u_int16_t lfs_activesb;		/* toggle between superblocks */
781 	daddr_t	  lfs_sbactive;		/* disk address of current sb write */
782 	struct vnode *lfs_flushvp;	/* vnode being flushed */
783 	struct vnode *lfs_unlockvp;	/* being inactivated in lfs_segunlock */
784 	u_int32_t lfs_diropwait;	/* # procs waiting on dirop flush */
785 	size_t lfs_devbsize;		/* Device block size */
786 	size_t lfs_devbshift;		/* Device block shift */
787 	struct lock lfs_fraglock;
788 	struct lock lfs_iflock;		/* Ifile lock */
789 	pid_t lfs_rfpid;		/* Process ID of roll-forward agent */
790 	int	  lfs_nadirop;		/* number of active dirop nodes */
791 	long	  lfs_ravail;		/* blocks pre-reserved for writing */
792 	long	  lfs_favail;		/* blocks pre-reserved for writing */
793 	res_t *lfs_resblk;		/* Reserved memory for pageout */
794 	TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */
795 	TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */
796 #define LFS_RESHASH_WIDTH 17
797 	LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH];
798 	int	  lfs_pdflush;		 /* pagedaemon wants us to flush */
799 	u_int32_t **lfs_suflags;	/* Segment use flags */
800 #ifdef _KERNEL
801 	struct pool lfs_clpool;		/* Pool for struct lfs_cluster */
802 	struct pool lfs_bpppool;	/* Pool for bpp */
803 	struct pool lfs_segpool;	/* Pool for struct segment */
804 #endif /* _KERNEL */
805 #define LFS_MAX_CLEANIND 64
806 	int32_t  lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */
807 	int 	 lfs_cleanind;	/* Index into intervals */
808 	struct simplelock lfs_interlock;  /* lock for lfs_seglock */
809 	int lfs_sleepers;		/* # procs sleeping this fs */
810 	int lfs_pages;			/* dirty pages blaming this fs */
811 };
812 
813 /* NINDIR is the number of indirects in a file system block. */
814 #define	NINDIR(fs)	((fs)->lfs_nindir)
815 
816 /* INOPB is the number of inodes in a secondary storage block. */
817 #define	INOPB(fs)	((fs)->lfs_inopb)
818 /* INOPF is the number of inodes in a fragment. */
819 #define INOPF(fs)	((fs)->lfs_inopf)
820 
821 #define	blksize(fs, ip, lbn) \
822 	(((lbn) >= NDADDR || (ip)->i_ffs1_size >= ((lbn) + 1) << (fs)->lfs_bshift) \
823 	    ? (fs)->lfs_bsize \
824 	    : (fragroundup(fs, blkoff(fs, (ip)->i_ffs1_size))))
825 #define	blkoff(fs, loc)		((int)((loc) & (fs)->lfs_bmask))
826 #define fragoff(fs, loc)    /* calculates (loc % fs->lfs_fsize) */ \
827     ((int)((loc) & (fs)->lfs_ffmask))
828 #define	fsbtodb(fs, b)		((b) << (fs)->lfs_fsbtodb)
829 #define	dbtofsb(fs, b)		((b) >> (fs)->lfs_fsbtodb)
830 #define fragstodb(fs, b)	((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift))
831 #define dbtofrags(fs, b)	((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift))
832 #define	lblkno(fs, loc)		((loc) >> (fs)->lfs_bshift)
833 #define	lblktosize(fs, blk)	((blk) << (fs)->lfs_bshift)
834 /* Same as above, but named like dbtob(), btodb() */
835 #define fsbtob(fs, b)		((b) << ((fs)->lfs_bshift - \
836 				(fs)->lfs_blktodb + (fs)->lfs_fsbtodb))
837 #define btofsb(fs, b)		((b) >> ((fs)->lfs_bshift - \
838 				(fs)->lfs_blktodb + (fs)->lfs_fsbtodb))
839 #define fsbtofrags(fs, b)	((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \
840 				(fs)->lfs_fsbtodb))
841 #define fragstofsb(fs, b)	((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \
842 				(fs)->lfs_fsbtodb))
843 #define btofrags(fs, b)		((b) >> (fs)->lfs_ffshift)
844 #define numfrags(fs, loc)	/* calculates (loc / fs->lfs_fsize) */	\
845 	((loc) >> (fs)->lfs_ffshift)
846 #define blkroundup(fs, size)	/* calculates roundup(size, fs->lfs_bsize) */ \
847 	((off_t)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask)))
848 #define fragroundup(fs, size)	/* calculates roundup(size, fs->lfs_fsize) */ \
849 	((off_t)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask)))
850 #define fragstoblks(fs, frags)	/* calculates (frags / fs->lfs_frag) */	\
851 	((frags) >> (fs)->lfs_fbshift)
852 #define blkstofrags(fs, blks)	/* calculates (blks * fs->lfs_frag) */	\
853 	((blks) << (fs)->lfs_fbshift)
854 #define fragnum(fs, fsb)	/* calculates (fsb % fs->lfs_frag) */	\
855 	((fsb) & ((fs)->lfs_frag - 1))
856 #define blknum(fs, fsb)		/* calculates rounddown(fsb, fs->lfs_frag) */ \
857 	((fsb) &~ ((fs)->lfs_frag - 1))
858 #define dblksize(fs, dp, lbn) \
859 	(((lbn) >= NDADDR || (dp)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\
860 	    ? (fs)->lfs_bsize \
861 	    : (fragroundup(fs, blkoff(fs, (dp)->di_size))))
862 
863 #define	segsize(fs)	((fs)->lfs_version == 1 ?	     		\
864 			   lblktosize((fs), (fs)->lfs_ssize) :		\
865 			   (fs)->lfs_ssize)
866 #define segtod(fs, seg) (((fs)->lfs_version == 1     ?	     		\
867 			   (fs)->lfs_ssize << (fs)->lfs_blktodb :	\
868 			   btofsb((fs), (fs)->lfs_ssize)) * (seg))
869 #define	dtosn(fs, daddr)	/* block address to segment number */	\
870 	((uint32_t)(((daddr) - (fs)->lfs_start) / segtod((fs), 1)))
871 #define sntod(fs, sn)		/* segment number to disk address */	\
872 	((daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start))
873 
874 /*
875  * Structures used by lfs_bmapv and lfs_markv to communicate information
876  * about inodes and data blocks.
877  */
878 typedef struct block_info {
879 	u_int32_t bi_inode;		/* inode # */
880 	int32_t	bi_lbn;			/* logical block w/in file */
881 	int32_t	bi_daddr;		/* disk address of block */
882 	u_int64_t bi_segcreate;		/* origin segment create time */
883 	int	bi_version;		/* file version number */
884 	void	*bi_bp;			/* data buffer */
885 	int	bi_size;		/* size of the block (if fragment) */
886 } BLOCK_INFO;
887 
888 /* Compatibility for 1.5 binaries */
889 typedef struct block_info_15 {
890 	u_int32_t bi_inode;		/* inode # */
891 	int32_t	bi_lbn;			/* logical block w/in file */
892 	int32_t	bi_daddr;		/* disk address of block */
893 	u_int32_t bi_segcreate;		/* origin segment create time */
894 	int	bi_version;		/* file version number */
895 	void	*bi_bp;			/* data buffer */
896 	int	bi_size;		/* size of the block (if fragment) */
897 } BLOCK_INFO_15;
898 
899 /* In-memory description of a segment about to be written. */
900 struct segment {
901 	struct lfs	 *fs;		/* file system pointer */
902 	struct buf	**bpp;		/* pointer to buffer array */
903 	struct buf	**cbpp;		/* pointer to next available bp */
904 	struct buf	**start_bpp;	/* pointer to first bp in this set */
905 	struct buf	 *ibp;		/* buffer pointer to inode page */
906 	struct ufs1_dinode    *idp;          /* pointer to ifile dinode */
907 	struct finfo	 *fip;		/* current fileinfo pointer */
908 	struct vnode	 *vp;		/* vnode being gathered */
909 	void	 *segsum;		/* segment summary info */
910 	u_int32_t ninodes;		/* number of inodes in this segment */
911 	int32_t seg_bytes_left;		/* bytes left in segment */
912 	int32_t sum_bytes_left;		/* bytes left in summary block */
913 	u_int32_t seg_number;		/* number of this segment */
914 	int32_t *start_lbp;		/* beginning lbn for this set */
915 
916 #define	SEGM_CKP	0x01		/* doing a checkpoint */
917 #define	SEGM_CLEAN	0x02		/* cleaner call; don't sort */
918 #define	SEGM_SYNC	0x04		/* wait for segment */
919 #define	SEGM_PROT	0x08		/* don't inactivate at segunlock */
920 #define SEGM_PAGEDAEMON	0x10		/* pagedaemon called us */
921 #define SEGM_WRITERD	0x20		/* LFS writed called us */
922 #define SEGM_FORCE_CKP	0x40		/* Force checkpoint right away */
923 	u_int16_t seg_flags;		/* run-time flags for this segment */
924 	u_int32_t seg_iocount;		/* number of ios pending */
925 	int	  ndupino;		/* number of duplicate inodes */
926 };
927 
928 #ifdef _KERNEL
929 struct lfs_cluster {
930 	size_t bufsize;	       /* Size of kept data */
931 	struct buf **bpp;      /* Array of kept buffers */
932 	int bufcount;	       /* Number of kept buffers */
933 #define LFS_CL_MALLOC	0x00000001
934 #define LFS_CL_SHIFT	0x00000002
935 #define LFS_CL_SYNC	0x00000004
936 	u_int32_t flags;       /* Flags */
937 	struct lfs *fs;	       /* LFS that this belongs to */
938 	struct segment *seg;   /* Segment structure, for LFS_CL_SYNC */
939 };
940 
941 /*
942  * Splay tree containing block numbers allocated through lfs_balloc.
943  */
944 struct lbnentry {
945 	SPLAY_ENTRY(lbnentry) entry;
946 	daddr_t lbn;
947 };
948 #endif /* _KERNEL */
949 
950 /*
951  * LFS inode extensions.
952  */
953 struct lfs_inode_ext {
954 	off_t	  lfs_osize;		/* size of file on disk */
955 	u_int32_t lfs_effnblocks;  /* number of blocks when i/o completes */
956 	size_t	  lfs_fragsize[NDADDR]; /* size of on-disk direct blocks */
957 	TAILQ_ENTRY(inode) lfs_dchain;  /* Dirop chain. */
958 	TAILQ_ENTRY(inode) lfs_pchain;  /* Paging chain. */
959 #define LFSI_NO_GOP_WRITE 0x01
960 	u_int32_t lfs_iflags;           /* Inode flags */
961 	daddr_t   lfs_hiblk;		/* Highest lbn held by inode */
962 #ifdef _KERNEL
963 	SPLAY_HEAD(lfs_splay, lbnentry) lfs_lbtree; /* Tree of balloc'd lbns */
964 	int	  lfs_nbtree;		/* Size of tree */
965 #endif
966 };
967 #define i_lfs_osize		inode_ext.lfs->lfs_osize
968 #define i_lfs_effnblks		inode_ext.lfs->lfs_effnblocks
969 #define i_lfs_fragsize		inode_ext.lfs->lfs_fragsize
970 #define i_lfs_dchain		inode_ext.lfs->lfs_dchain
971 #define i_lfs_iflags		inode_ext.lfs->lfs_iflags
972 #define i_lfs_hiblk		inode_ext.lfs->lfs_hiblk
973 #define i_lfs_lbtree		inode_ext.lfs->lfs_lbtree
974 #define i_lfs_nbtree		inode_ext.lfs->lfs_nbtree
975 
976 /*
977  * Macros for determining free space on the disk, with the variable metadata
978  * of segment summaries and inode blocks taken into account.
979  */
980 /* Estimate number of clean blocks not available for writing */
981 #define LFS_EST_CMETA(F) (int32_t)((((F)->lfs_dmeta *			     \
982 				     (int64_t)(F)->lfs_nclean) /	     \
983 				      ((F)->lfs_nseg - (F)->lfs_nclean)))
984 
985 /* Estimate total size of the disk not including metadata */
986 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F))
987 
988 /* Estimate number of blocks actually available for writing */
989 #define LFS_EST_BFREE(F) ((F)->lfs_bfree > LFS_EST_CMETA(F) + (F)->lfs_dmeta ? (F)->lfs_bfree - LFS_EST_CMETA(F) - (F)->lfs_dmeta : 0)
990 
991 /* Amount of non-meta space not available to mortal man */
992 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) *			     \
993 				   (u_int64_t)(F)->lfs_minfree) /	     \
994 				  100)
995 
996 /* Can credential C write BB blocks */
997 #define ISSPACE(F, BB, C)						\
998 	((((C) == NOCRED || (C)->cr_uid == 0) &&			\
999 	  LFS_EST_BFREE(F) >= (BB)) ||					\
1000 	 ((C)->cr_uid != 0 && IS_FREESPACE(F, BB)))
1001 
1002 /* Can an ordinary user write BB blocks */
1003 #define IS_FREESPACE(F, BB)						\
1004 	  (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F))
1005 
1006 /*
1007  * The minimum number of blocks to create a new inode.  This is:
1008  * directory direct block (1) + NIADDR indirect blocks + inode block (1) +
1009  * ifile direct block (1) + NIADDR indirect blocks = 3 + 2 * NIADDR blocks.
1010  */
1011 #define LFS_NRESERVE(F) (btofsb((F), (2 * NIADDR + 3) << (F)->lfs_bshift))
1012 
1013 /* Statistics Counters */
1014 struct lfs_stats {	/* Must match sysctl list in lfs_vfsops.h ! */
1015 	u_int	segsused;
1016 	u_int	psegwrites;
1017 	u_int	psyncwrites;
1018 	u_int	pcleanwrites;
1019 	u_int	blocktot;
1020 	u_int	cleanblocks;
1021 	u_int	ncheckpoints;
1022 	u_int	nwrites;
1023 	u_int	nsync_writes;
1024 	u_int	wait_exceeded;
1025 	u_int	write_exceeded;
1026 	u_int	flush_invoked;
1027 	u_int	vflush_invoked;
1028 	u_int	clean_inlocked;
1029 	u_int	clean_vnlocked;
1030 	u_int   segs_reclaimed;
1031 };
1032 #ifdef _KERNEL
1033 extern struct lfs_stats lfs_stats;
1034 #endif
1035 
1036 /* Fcntls to take the place of the lfs syscalls */
1037 struct lfs_fcntl_markv {
1038 	BLOCK_INFO *blkiov;	/* blocks to relocate */
1039 	int blkcnt;		/* number of blocks */
1040 };
1041 
1042 #define LFCNSEGWAITALL	 _FCNR_FSPRIV('L', 0, struct timeval)
1043 #define LFCNSEGWAIT	 _FCNR_FSPRIV('L', 1, struct timeval)
1044 #define LFCNBMAPV	_FCNRW_FSPRIV('L', 2, struct lfs_fcntl_markv)
1045 #define LFCNMARKV	_FCNRW_FSPRIV('L', 3, struct lfs_fcntl_markv)
1046 #define LFCNRECLAIM	 _FCNO_FSPRIV('L', 4)
1047 #define LFCNIFILEFH	 _FCNW_FSPRIV('L', 5, struct fhandle)
1048 #define LFCNREWIND       _FCNR_FSPRIV('L', 6, int)
1049 #define LFCNINVAL        _FCNR_FSPRIV('L', 7, int)
1050 #define LFCNRESIZE       _FCNR_FSPRIV('L', 8, int)
1051 /* Compat for NetBSD 2.x bug */
1052 #define LFCNSEGWAITALL_COMPAT	 _FCNW_FSPRIV('L', 0, struct timeval)
1053 #define LFCNSEGWAIT_COMPAT	 _FCNW_FSPRIV('L', 1, struct timeval)
1054 
1055 #ifdef _KERNEL
1056 /* XXX MP */
1057 #define	LFS_SEGLOCK_HELD(fs) \
1058 	((fs)->lfs_seglock != 0 && (fs)->lfs_lockpid == curproc->p_pid)
1059 #endif /* _KERNEL */
1060 
1061 /* Debug segment lock */
1062 #ifdef notyet
1063 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs))
1064 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs))
1065 # define ASSERT_DUNNO_SEGLOCK(fs)
1066 # define ASSERT_MAYBE_SEGLOCK(fs)
1067 #else /* !notyet */
1068 # define ASSERT_DUNNO_SEGLOCK(fs) \
1069 	DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \
1070 		LFS_SEGLOCK_HELD(fs)))
1071 # define ASSERT_SEGLOCK(fs) do {					\
1072 	if (!LFS_SEGLOCK_HELD(fs)) {					\
1073 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \
1074 	}								\
1075 } while(0)
1076 # define ASSERT_NO_SEGLOCK(fs) do {					\
1077 	if (LFS_SEGLOCK_HELD(fs)) {					\
1078 		DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \
1079 	}								\
1080 } while(0)
1081 # define ASSERT_MAYBE_SEGLOCK(x)
1082 #endif /* !notyet */
1083 
1084 __BEGIN_DECLS
1085 void lfs_itimes(struct inode *, const struct timespec *,
1086     const struct timespec *, const struct timespec *);
1087 __END_DECLS
1088 
1089 #endif /* !_UFS_LFS_LFS_H_ */
1090