1 /* $NetBSD: lfs.h,v 1.133 2010/02/16 23:20:30 mlelstv 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 * 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 70 /* 71 * Compile-time options for LFS. 72 */ 73 #define LFS_IFIND_RETRIES 16 74 #define LFS_LOGLENGTH 1024 /* size of debugging log */ 75 #define LFS_MAX_ACTIVE 10 /* Dirty segments before ckp forced */ 76 77 /* 78 * Fixed filesystem layout parameters 79 */ 80 #define LFS_LABELPAD 8192 /* LFS label size */ 81 #define LFS_SBPAD 8192 /* LFS superblock size */ 82 83 #define LFS_UNUSED_INUM 0 /* 0: out of band inode number */ 84 #define LFS_IFILE_INUM 1 /* 1: IFILE inode number */ 85 /* 2: Root inode number */ 86 #define LOSTFOUNDINO 3 /* 3: lost+found inode number */ 87 #define LFS_FIRST_INUM 4 /* 4: first free inode number */ 88 89 #define LFS_V1_SUMMARY_SIZE 512 /* V1 fixed summary size */ 90 #define LFS_DFL_SUMMARY_SIZE 512 /* Default summary size */ 91 92 #define LFS_MAX_DADDR 0x7fffffff /* Highest addressable fsb */ 93 94 #define LFS_MAXNAMLEN 255 /* maximum name length in a dir */ 95 96 /* Adjustable filesystem parameters */ 97 #define MIN_FREE_SEGS 20 98 #define MIN_RESV_SEGS 15 99 #ifndef LFS_ATIME_IFILE 100 # define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */ 101 #endif 102 #define LFS_MARKV_MAXBLKCNT 65536 /* Max block count for lfs_markv() */ 103 104 /* Misc. definitions */ 105 #define BW_CLEAN 1 /* Flag for lfs_bwrite_ext() */ 106 #define PG_DELWRI PG_PAGER1 /* Local def for delayed pageout */ 107 108 /* Resource limits */ 109 #define LFS_MAX_RESOURCE(x, u) (((x) >> 2) - 10 * (u)) 110 #define LFS_WAIT_RESOURCE(x, u) (((x) >> 1) - ((x) >> 3) - 10 * (u)) 111 #define LFS_INVERSE_MAX_RESOURCE(x, u) (((x) + 10 * (u)) << 2) 112 #define LFS_MAX_BUFS LFS_MAX_RESOURCE(nbuf, 1) 113 #define LFS_WAIT_BUFS LFS_WAIT_RESOURCE(nbuf, 1) 114 #define LFS_INVERSE_MAX_BUFS(n) LFS_INVERSE_MAX_RESOURCE(n, 1) 115 #define LFS_MAX_BYTES LFS_MAX_RESOURCE(bufmem_lowater, PAGE_SIZE) 116 #define LFS_INVERSE_MAX_BYTES(n) LFS_INVERSE_MAX_RESOURCE(n, PAGE_SIZE) 117 #define LFS_WAIT_BYTES LFS_WAIT_RESOURCE(bufmem_lowater, PAGE_SIZE) 118 #define LFS_MAX_DIROP ((desiredvnodes >> 2) + (desiredvnodes >> 3)) 119 #define SIZEOF_DIROP(fs) (2 * ((fs)->lfs_bsize + DINODE1_SIZE)) 120 #define LFS_MAX_FSDIROP(fs) \ 121 ((fs)->lfs_nclean <= (fs)->lfs_resvseg ? 0 : \ 122 (((fs)->lfs_nclean - (fs)->lfs_resvseg) * (fs)->lfs_ssize) / \ 123 (2 * SIZEOF_DIROP(fs))) 124 #define LFS_MAX_PAGES lfs_max_pages() 125 #define LFS_WAIT_PAGES lfs_wait_pages() 126 #define LFS_BUFWAIT 2 /* How long to wait if over *_WAIT_* */ 127 128 #ifdef _KERNEL 129 int lfs_wait_pages(void); 130 int lfs_max_pages(void); 131 #endif /* _KERNEL */ 132 133 /* How starved can we be before we start holding back page writes */ 134 #define LFS_STARVED_FOR_SEGS(fs) ((fs)->lfs_nclean < (fs)->lfs_resvseg) 135 136 /* 137 * Reserved blocks for lfs_malloc 138 */ 139 140 /* Structure to keep reserved blocks */ 141 typedef struct lfs_res_blk { 142 void *p; 143 LIST_ENTRY(lfs_res_blk) res; 144 int size; 145 char inuse; 146 } res_t; 147 148 /* Types for lfs_newbuf and lfs_malloc */ 149 #define LFS_NB_UNKNOWN -1 150 #define LFS_NB_SUMMARY 0 151 #define LFS_NB_SBLOCK 1 152 #define LFS_NB_IBLOCK 2 153 #define LFS_NB_CLUSTER 3 154 #define LFS_NB_CLEAN 4 155 #define LFS_NB_BLKIOV 5 156 #define LFS_NB_COUNT 6 /* always last */ 157 158 /* Number of reserved memory blocks of each type */ 159 #define LFS_N_SUMMARIES 2 160 #define LFS_N_SBLOCKS 1 /* Always 1, to throttle superblock writes */ 161 #define LFS_N_IBLOCKS 16 /* In theory ssize/bsize; in practice around 2 */ 162 #define LFS_N_CLUSTERS 16 /* In theory ssize/MAXPHYS */ 163 #define LFS_N_CLEAN 0 164 #define LFS_N_BLKIOV 1 165 166 /* Total count of "large" (non-pool) types */ 167 #define LFS_N_TOTAL (LFS_N_SUMMARIES + LFS_N_SBLOCKS + LFS_N_IBLOCKS + \ 168 LFS_N_CLUSTERS + LFS_N_CLEAN + LFS_N_BLKIOV) 169 170 /* Counts for pool types */ 171 #define LFS_N_CL LFS_N_CLUSTERS 172 #define LFS_N_BPP 2 173 #define LFS_N_SEG 2 174 175 /* 176 * "struct buf" associated definitions 177 */ 178 179 /* Unassigned disk addresses. */ 180 #define UNASSIGNED -1 181 #define UNWRITTEN -2 182 183 /* Unused logical block number */ 184 #define LFS_UNUSED_LBN -1 185 186 /* Determine if a buffer belongs to the ifile */ 187 #define IS_IFILE(bp) (VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM) 188 189 # define LFS_LOCK_BUF(bp) do { \ 190 if (((bp)->b_flags & B_LOCKED) == 0 && bp->b_iodone == NULL) { \ 191 mutex_enter(&lfs_lock); \ 192 ++locked_queue_count; \ 193 locked_queue_bytes += bp->b_bufsize; \ 194 mutex_exit(&lfs_lock); \ 195 } \ 196 (bp)->b_flags |= B_LOCKED; \ 197 } while (0) 198 199 # define LFS_UNLOCK_BUF(bp) do { \ 200 if (((bp)->b_flags & B_LOCKED) != 0 && bp->b_iodone == NULL) { \ 201 mutex_enter(&lfs_lock); \ 202 --locked_queue_count; \ 203 locked_queue_bytes -= bp->b_bufsize; \ 204 if (locked_queue_count < LFS_WAIT_BUFS && \ 205 locked_queue_bytes < LFS_WAIT_BYTES) \ 206 cv_broadcast(&locked_queue_cv); \ 207 mutex_exit(&lfs_lock); \ 208 } \ 209 (bp)->b_flags &= ~B_LOCKED; \ 210 } while (0) 211 212 #ifdef _KERNEL 213 214 extern u_long bufmem_lowater, bufmem_hiwater; /* XXX */ 215 216 # define LFS_IS_MALLOC_BUF(bp) ((bp)->b_iodone == lfs_callback) 217 218 # ifdef DEBUG 219 # define LFS_DEBUG_COUNTLOCKED(m) do { \ 220 if (lfs_debug_log_subsys[DLOG_LLIST]) { \ 221 lfs_countlocked(&locked_queue_count, &locked_queue_bytes, (m)); \ 222 cv_broadcast(&locked_queue_cv); \ 223 } \ 224 } while (0) 225 # else 226 # define LFS_DEBUG_COUNTLOCKED(m) 227 # endif 228 229 /* log for debugging writes to the Ifile */ 230 # ifdef DEBUG 231 struct lfs_log_entry { 232 const char *op; 233 const char *file; 234 int pid; 235 int line; 236 daddr_t block; 237 unsigned long flags; 238 }; 239 extern int lfs_lognum; 240 extern struct lfs_log_entry lfs_log[LFS_LOGLENGTH]; 241 # define LFS_BWRITE_LOG(bp) lfs_bwrite_log((bp), __FILE__, __LINE__) 242 # define LFS_ENTER_LOG(theop, thefile, theline, lbn, theflags, thepid) do {\ 243 int _s; \ 244 \ 245 mutex_enter(&lfs_lock); \ 246 _s = splbio(); \ 247 lfs_log[lfs_lognum].op = theop; \ 248 lfs_log[lfs_lognum].file = thefile; \ 249 lfs_log[lfs_lognum].line = (theline); \ 250 lfs_log[lfs_lognum].pid = (thepid); \ 251 lfs_log[lfs_lognum].block = (lbn); \ 252 lfs_log[lfs_lognum].flags = (theflags); \ 253 lfs_lognum = (lfs_lognum + 1) % LFS_LOGLENGTH; \ 254 splx(_s); \ 255 mutex_exit(&lfs_lock); \ 256 } while (0) 257 258 # define LFS_BCLEAN_LOG(fs, bp) do { \ 259 if ((bp)->b_vp == (fs)->lfs_ivnode) \ 260 LFS_ENTER_LOG("clear", __FILE__, __LINE__, \ 261 bp->b_lblkno, bp->b_flags, curproc->p_pid);\ 262 } while (0) 263 264 /* Must match list in lfs_vfsops.c ! */ 265 # define DLOG_RF 0 /* roll forward */ 266 # define DLOG_ALLOC 1 /* inode alloc */ 267 # define DLOG_AVAIL 2 /* lfs_{,r,f}avail */ 268 # define DLOG_FLUSH 3 /* flush */ 269 # define DLOG_LLIST 4 /* locked list accounting */ 270 # define DLOG_WVNODE 5 /* vflush/writevnodes verbose */ 271 # define DLOG_VNODE 6 /* vflush/writevnodes */ 272 # define DLOG_SEG 7 /* segwrite */ 273 # define DLOG_SU 8 /* seguse accounting */ 274 # define DLOG_CLEAN 9 /* cleaner routines */ 275 # define DLOG_MOUNT 10 /* mount/unmount */ 276 # define DLOG_PAGE 11 /* putpages/gop_write */ 277 # define DLOG_DIROP 12 /* dirop accounting */ 278 # define DLOG_MALLOC 13 /* lfs_malloc accounting */ 279 # define DLOG_MAX 14 /* The terminator */ 280 # define DLOG(a) lfs_debug_log a 281 # else /* ! DEBUG */ 282 # define LFS_BCLEAN_LOG(fs, bp) 283 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)) 284 # define DLOG(a) 285 # endif /* ! DEBUG */ 286 #else /* ! _KERNEL */ 287 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)) 288 #endif /* _KERNEL */ 289 290 #ifdef _KERNEL 291 /* Filehandle structure for exported LFSes */ 292 struct lfid { 293 struct ufid lfid_ufid; 294 #define lfid_len lfid_ufid.ufid_len 295 #define lfid_ino lfid_ufid.ufid_ino 296 #define lfid_gen lfid_ufid.ufid_gen 297 uint32_t lfid_ident; 298 }; 299 #endif /* _KERNEL */ 300 301 /* 302 * "struct inode" associated definitions 303 */ 304 305 /* Address calculations for metadata located in the inode */ 306 #define S_INDIR(fs) -NDADDR 307 #define D_INDIR(fs) (S_INDIR(fs) - NINDIR(fs) - 1) 308 #define T_INDIR(fs) (D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1) 309 310 /* For convenience */ 311 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING) 312 313 #define LFS_SET_UINO(ip, flags) do { \ 314 if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED)) \ 315 ++(ip)->i_lfs->lfs_uinodes; \ 316 if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING)) \ 317 ++(ip)->i_lfs->lfs_uinodes; \ 318 if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED)) \ 319 ++(ip)->i_lfs->lfs_uinodes; \ 320 (ip)->i_flag |= (flags); \ 321 } while (0) 322 323 #define LFS_CLR_UINO(ip, flags) do { \ 324 if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED)) \ 325 --(ip)->i_lfs->lfs_uinodes; \ 326 if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING)) \ 327 --(ip)->i_lfs->lfs_uinodes; \ 328 if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED)) \ 329 --(ip)->i_lfs->lfs_uinodes; \ 330 (ip)->i_flag &= ~(flags); \ 331 if ((ip)->i_lfs->lfs_uinodes < 0) { \ 332 panic("lfs_uinodes < 0"); \ 333 } \ 334 } while (0) 335 336 #define LFS_ITIMES(ip, acc, mod, cre) \ 337 while ((ip)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY)) \ 338 lfs_itimes(ip, acc, mod, cre) 339 340 /* 341 * "struct vnode" associated definitions 342 */ 343 344 /* Heuristic emptiness measure */ 345 #define VPISEMPTY(vp) (LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 346 !(vp->v_type == VREG && (vp)->v_iflag & VI_ONWORKLST) &&\ 347 VTOI(vp)->i_lfs_nbtree == 0) 348 349 #define WRITEINPROG(vp) ((vp)->v_numoutput > 0 || \ 350 (!LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 351 !(VTOI(vp)->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING)))) 352 353 354 /* 355 * On-disk and in-memory checkpoint segment usage structure. 356 */ 357 typedef struct segusage SEGUSE; 358 struct segusage { 359 u_int32_t su_nbytes; /* 0: number of live bytes */ 360 u_int32_t su_olastmod; /* 4: SEGUSE last modified timestamp */ 361 u_int16_t su_nsums; /* 8: number of summaries in segment */ 362 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 363 364 #define SEGUSE_ACTIVE 0x01 /* segment currently being written */ 365 #define SEGUSE_DIRTY 0x02 /* segment has data in it */ 366 #define SEGUSE_SUPERBLOCK 0x04 /* segment contains a superblock */ 367 #define SEGUSE_ERROR 0x08 /* cleaner: do not clean segment */ 368 #define SEGUSE_EMPTY 0x10 /* segment is empty */ 369 #define SEGUSE_INVAL 0x20 /* segment is invalid */ 370 u_int32_t su_flags; /* 12: segment flags */ 371 u_int64_t su_lastmod; /* 16: last modified timestamp */ 372 }; 373 374 typedef struct segusage_v1 SEGUSE_V1; 375 struct segusage_v1 { 376 u_int32_t su_nbytes; /* 0: number of live bytes */ 377 u_int32_t su_lastmod; /* 4: SEGUSE last modified timestamp */ 378 u_int16_t su_nsums; /* 8: number of summaries in segment */ 379 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 380 u_int32_t su_flags; /* 12: segment flags */ 381 }; 382 383 #define SEGUPB(fs) (fs->lfs_sepb) 384 #define SEGTABSIZE_SU(fs) \ 385 (((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb) 386 387 #ifdef _KERNEL 388 # define SHARE_IFLOCK(F) \ 389 do { \ 390 rw_enter(&(F)->lfs_iflock, RW_READER); \ 391 } while(0) 392 # define UNSHARE_IFLOCK(F) \ 393 do { \ 394 rw_exit(&(F)->lfs_iflock); \ 395 } while(0) 396 #else /* ! _KERNEL */ 397 # define SHARE_IFLOCK(F) 398 # define UNSHARE_IFLOCK(F) 399 #endif /* ! _KERNEL */ 400 401 /* Read in the block with a specific segment usage entry from the ifile. */ 402 #define LFS_SEGENTRY(SP, F, IN, BP) do { \ 403 int _e; \ 404 SHARE_IFLOCK(F); \ 405 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 406 if ((_e = bread((F)->lfs_ivnode, \ 407 ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz, \ 408 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 409 panic("lfs: ifile read: %d", _e); \ 410 if ((F)->lfs_version == 1) \ 411 (SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data + \ 412 ((IN) & ((F)->lfs_sepb - 1))); \ 413 else \ 414 (SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb); \ 415 UNSHARE_IFLOCK(F); \ 416 } while (0) 417 418 #define LFS_WRITESEGENTRY(SP, F, IN, BP) do { \ 419 if ((SP)->su_nbytes == 0) \ 420 (SP)->su_flags |= SEGUSE_EMPTY; \ 421 else \ 422 (SP)->su_flags &= ~SEGUSE_EMPTY; \ 423 (F)->lfs_suflags[(F)->lfs_activesb][(IN)] = (SP)->su_flags; \ 424 LFS_BWRITE_LOG(BP); \ 425 } while (0) 426 427 /* 428 * On-disk file information. One per file with data blocks in the segment. 429 */ 430 typedef struct finfo FINFO; 431 struct finfo { 432 u_int32_t fi_nblocks; /* number of blocks */ 433 u_int32_t fi_version; /* version number */ 434 u_int32_t fi_ino; /* inode number */ 435 u_int32_t fi_lastlength; /* length of last block in array */ 436 int32_t fi_blocks[1]; /* array of logical block numbers */ 437 }; 438 /* sizeof FINFO except fi_blocks */ 439 #define FINFOSIZE (sizeof(FINFO) - sizeof(int32_t)) 440 441 /* 442 * Index file inode entries. 443 */ 444 typedef struct ifile IFILE; 445 struct ifile { 446 u_int32_t if_version; /* inode version number */ 447 #define LFS_UNUSED_DADDR 0 /* out-of-band daddr */ 448 int32_t if_daddr; /* inode disk address */ 449 #define LFS_ORPHAN_NEXTFREE (~(u_int32_t)0) /* indicate orphaned file */ 450 u_int32_t if_nextfree; /* next-unallocated inode */ 451 u_int32_t if_atime_sec; /* Last access time, seconds */ 452 u_int32_t if_atime_nsec; /* and nanoseconds */ 453 }; 454 455 typedef struct ifile_v1 IFILE_V1; 456 struct ifile_v1 { 457 u_int32_t if_version; /* inode version number */ 458 int32_t if_daddr; /* inode disk address */ 459 u_int32_t if_nextfree; /* next-unallocated inode */ 460 #if LFS_ATIME_IFILE 461 struct timespec if_atime; /* Last access time */ 462 #endif 463 }; 464 465 /* 466 * LFSv1 compatibility code is not allowed to touch if_atime, since it 467 * may not be mapped! 468 */ 469 /* Read in the block with a specific inode from the ifile. */ 470 #define LFS_IENTRY(IP, F, IN, BP) do { \ 471 int _e; \ 472 SHARE_IFLOCK(F); \ 473 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 474 if ((_e = bread((F)->lfs_ivnode, \ 475 (IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz, \ 476 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 477 panic("lfs: ifile ino %d read %d", (int)(IN), _e); \ 478 if ((F)->lfs_version == 1) \ 479 (IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data + \ 480 (IN) % (F)->lfs_ifpb); \ 481 else \ 482 (IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb; \ 483 UNSHARE_IFLOCK(F); \ 484 } while (0) 485 486 /* 487 * Cleaner information structure. This resides in the ifile and is used 488 * to pass information from the kernel to the cleaner. 489 */ 490 typedef struct _cleanerinfo { 491 u_int32_t clean; /* number of clean segments */ 492 u_int32_t dirty; /* number of dirty segments */ 493 int32_t bfree; /* disk blocks free */ 494 int32_t avail; /* disk blocks available */ 495 u_int32_t free_head; /* head of the inode free list */ 496 u_int32_t free_tail; /* tail of the inode free list */ 497 #define LFS_CLEANER_MUST_CLEAN 0x01 498 u_int32_t flags; /* status word from the kernel */ 499 } CLEANERINFO; 500 501 #define CLEANSIZE_SU(fs) \ 502 ((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift) 503 504 /* Read in the block with the cleaner info from the ifile. */ 505 #define LFS_CLEANERINFO(CP, F, BP) do { \ 506 SHARE_IFLOCK(F); \ 507 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 508 if (bread((F)->lfs_ivnode, \ 509 (daddr_t)0, (F)->lfs_bsize, NOCRED, 0, &(BP))) \ 510 panic("lfs: ifile read"); \ 511 (CP) = (CLEANERINFO *)(BP)->b_data; \ 512 UNSHARE_IFLOCK(F); \ 513 } while (0) 514 515 /* 516 * Synchronize the Ifile cleaner info with current avail and bfree. 517 */ 518 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do { \ 519 mutex_enter(&lfs_lock); \ 520 if ((w) || (cip)->bfree != (fs)->lfs_bfree || \ 521 (cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail - \ 522 (fs)->lfs_favail) { \ 523 (cip)->bfree = (fs)->lfs_bfree; \ 524 (cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail - \ 525 (fs)->lfs_favail; \ 526 if (((bp)->b_flags & B_GATHERED) == 0) { \ 527 (fs)->lfs_flags |= LFS_IFDIRTY; \ 528 } \ 529 mutex_exit(&lfs_lock); \ 530 (void) LFS_BWRITE_LOG(bp); /* Ifile */ \ 531 } else { \ 532 mutex_exit(&lfs_lock); \ 533 brelse(bp, 0); \ 534 } \ 535 } while (0) 536 537 /* 538 * Get the head of the inode free list. 539 * Always called with the segment lock held. 540 */ 541 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do { \ 542 if ((FS)->lfs_version > 1) { \ 543 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 544 (FS)->lfs_freehd = (CIP)->free_head; \ 545 brelse(BP, 0); \ 546 } \ 547 *(FREEP) = (FS)->lfs_freehd; \ 548 } while (0) 549 550 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do { \ 551 (FS)->lfs_freehd = (VAL); \ 552 if ((FS)->lfs_version > 1) { \ 553 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 554 (CIP)->free_head = (VAL); \ 555 LFS_BWRITE_LOG(BP); \ 556 mutex_enter(&lfs_lock); \ 557 (FS)->lfs_flags |= LFS_IFDIRTY; \ 558 mutex_exit(&lfs_lock); \ 559 } \ 560 } while (0) 561 562 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do { \ 563 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 564 *(FREEP) = (CIP)->free_tail; \ 565 brelse(BP, 0); \ 566 } while (0) 567 568 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do { \ 569 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 570 (CIP)->free_tail = (VAL); \ 571 LFS_BWRITE_LOG(BP); \ 572 mutex_enter(&lfs_lock); \ 573 (FS)->lfs_flags |= LFS_IFDIRTY; \ 574 mutex_exit(&lfs_lock); \ 575 } while (0) 576 577 /* 578 * On-disk segment summary information 579 */ 580 typedef struct segsum_v1 SEGSUM_V1; 581 struct segsum_v1 { 582 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 583 u_int32_t ss_datasum; /* 4: check sum of data */ 584 u_int32_t ss_magic; /* 8: segment summary magic number */ 585 #define SS_MAGIC 0x061561 586 int32_t ss_next; /* 12: next segment */ 587 u_int32_t ss_create; /* 16: creation time stamp */ 588 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 589 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 590 591 #define SS_DIROP 0x01 /* segment begins a dirop */ 592 #define SS_CONT 0x02 /* more partials to finish this write*/ 593 #define SS_CLEAN 0x04 /* written by the cleaner */ 594 #define SS_RFW 0x08 /* written by the roll-forward agent */ 595 u_int16_t ss_flags; /* 24: used for directory operations */ 596 u_int16_t ss_pad; /* 26: extra space */ 597 /* FINFO's and inode daddr's... */ 598 }; 599 600 typedef struct segsum SEGSUM; 601 struct segsum { 602 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 603 u_int32_t ss_datasum; /* 4: check sum of data */ 604 u_int32_t ss_magic; /* 8: segment summary magic number */ 605 int32_t ss_next; /* 12: next segment */ 606 u_int32_t ss_ident; /* 16: roll-forward fsid */ 607 #define ss_ocreate ss_ident /* ident is where create was in v1 */ 608 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 609 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 610 u_int16_t ss_flags; /* 24: used for directory operations */ 611 u_int8_t ss_pad[6]; /* 26: extra space */ 612 u_int64_t ss_serial; /* 32: serial number */ 613 u_int64_t ss_create; /* 40: time stamp */ 614 /* FINFO's and inode daddr's... */ 615 }; 616 617 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM)) 618 619 620 /* 621 * On-disk super block. 622 */ 623 struct dlfs { 624 #define LFS_MAGIC 0x070162 625 u_int32_t dlfs_magic; /* 0: magic number */ 626 #define LFS_VERSION 2 627 u_int32_t dlfs_version; /* 4: version number */ 628 629 u_int32_t dlfs_size; /* 8: number of blocks in fs (v1) */ 630 /* number of frags in fs (v2) */ 631 u_int32_t dlfs_ssize; /* 12: number of blocks per segment (v1) */ 632 /* number of bytes per segment (v2) */ 633 u_int32_t dlfs_dsize; /* 16: number of disk blocks in fs */ 634 u_int32_t dlfs_bsize; /* 20: file system block size */ 635 u_int32_t dlfs_fsize; /* 24: size of frag blocks in fs */ 636 u_int32_t dlfs_frag; /* 28: number of frags in a block in fs */ 637 638 /* Checkpoint region. */ 639 u_int32_t dlfs_freehd; /* 32: start of the free list */ 640 int32_t dlfs_bfree; /* 36: number of free disk blocks */ 641 u_int32_t dlfs_nfiles; /* 40: number of allocated inodes */ 642 int32_t dlfs_avail; /* 44: blocks available for writing */ 643 int32_t dlfs_uinodes; /* 48: inodes in cache not yet on disk */ 644 int32_t dlfs_idaddr; /* 52: inode file disk address */ 645 u_int32_t dlfs_ifile; /* 56: inode file inode number */ 646 int32_t dlfs_lastseg; /* 60: address of last segment written */ 647 int32_t dlfs_nextseg; /* 64: address of next segment to write */ 648 int32_t dlfs_curseg; /* 68: current segment being written */ 649 int32_t dlfs_offset; /* 72: offset in curseg for next partial */ 650 int32_t dlfs_lastpseg; /* 76: address of last partial written */ 651 u_int32_t dlfs_inopf; /* 80: v1: time stamp; v2: inodes per frag */ 652 #define dlfs_otstamp dlfs_inopf 653 654 /* These are configuration parameters. */ 655 u_int32_t dlfs_minfree; /* 84: minimum percentage of free blocks */ 656 657 /* These fields can be computed from the others. */ 658 u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */ 659 u_int32_t dlfs_fsbpseg; /* 96: fsb per segment */ 660 u_int32_t dlfs_inopb; /* 100: inodes per block */ 661 u_int32_t dlfs_ifpb; /* 104: IFILE entries per block */ 662 u_int32_t dlfs_sepb; /* 108: SEGUSE entries per block */ 663 u_int32_t dlfs_nindir; /* 112: indirect pointers per block */ 664 u_int32_t dlfs_nseg; /* 116: number of segments */ 665 u_int32_t dlfs_nspf; /* 120: number of sectors per fragment */ 666 u_int32_t dlfs_cleansz; /* 124: cleaner info size in blocks */ 667 u_int32_t dlfs_segtabsz; /* 128: segment table size in blocks */ 668 u_int32_t dlfs_segmask; /* 132: calculate offset within a segment */ 669 u_int32_t dlfs_segshift; /* 136: fast mult/div for segments */ 670 u_int32_t dlfs_bshift; /* 140: calc block number from file offset */ 671 u_int32_t dlfs_ffshift; /* 144: fast mult/div for frag from file */ 672 u_int32_t dlfs_fbshift; /* 148: fast mult/div for frag from block */ 673 u_int64_t dlfs_bmask; /* 152: calc block offset from file offset */ 674 u_int64_t dlfs_ffmask; /* 160: calc frag offset from file offset */ 675 u_int64_t dlfs_fbmask; /* 168: calc frag offset from block offset */ 676 u_int32_t dlfs_blktodb; /* 176: blktodb and dbtoblk shift constant */ 677 u_int32_t dlfs_sushift; /* 180: fast mult/div for segusage table */ 678 679 int32_t dlfs_maxsymlinklen; /* 184: max length of an internal symlink */ 680 #define LFS_MIN_SBINTERVAL 5 /* minimum superblock segment spacing */ 681 #define LFS_MAXNUMSB 10 /* 188: superblock disk offsets */ 682 int32_t dlfs_sboffs[LFS_MAXNUMSB]; 683 684 u_int32_t dlfs_nclean; /* 228: Number of clean segments */ 685 u_char dlfs_fsmnt[MNAMELEN]; /* 232: name mounted on */ 686 #define LFS_PF_CLEAN 0x1 687 u_int16_t dlfs_pflags; /* 322: file system persistent flags */ 688 int32_t dlfs_dmeta; /* 324: total number of dirty summaries */ 689 u_int32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */ 690 u_int32_t dlfs_sumsize; /* 332: size of summary blocks */ 691 u_int64_t dlfs_serial; /* 336: serial number */ 692 u_int32_t dlfs_ibsize; /* 344: size of inode blocks */ 693 int32_t dlfs_start; /* 348: start of segment 0 */ 694 u_int64_t dlfs_tstamp; /* 352: time stamp */ 695 #define LFS_44INODEFMT 0 696 #define LFS_MAXINODEFMT 0 697 u_int32_t dlfs_inodefmt; /* 360: inode format version */ 698 u_int32_t dlfs_interleave; /* 364: segment interleave */ 699 u_int32_t dlfs_ident; /* 368: per-fs identifier */ 700 u_int32_t dlfs_fsbtodb; /* 372: fsbtodb abd dbtodsb shift constant */ 701 u_int32_t dlfs_resvseg; /* 376: segments reserved for the cleaner */ 702 int8_t dlfs_pad[128]; /* 380: round to 512 bytes */ 703 /* Checksum -- last valid disk field. */ 704 u_int32_t dlfs_cksum; /* 508: checksum for superblock checking */ 705 }; 706 707 /* Type used for the inode bitmap */ 708 typedef u_int32_t lfs_bm_t; 709 710 /* 711 * Linked list of segments whose byte count needs updating following a 712 * file truncation. 713 */ 714 struct segdelta { 715 long segnum; 716 size_t num; 717 LIST_ENTRY(segdelta) list; 718 }; 719 720 /* 721 * In-memory super block. 722 */ 723 struct lfs { 724 struct dlfs lfs_dlfs; /* on-disk parameters */ 725 #define lfs_magic lfs_dlfs.dlfs_magic 726 #define lfs_version lfs_dlfs.dlfs_version 727 #define lfs_size lfs_dlfs.dlfs_size 728 #define lfs_ssize lfs_dlfs.dlfs_ssize 729 #define lfs_dsize lfs_dlfs.dlfs_dsize 730 #define lfs_bsize lfs_dlfs.dlfs_bsize 731 #define lfs_fsize lfs_dlfs.dlfs_fsize 732 #define lfs_frag lfs_dlfs.dlfs_frag 733 #define lfs_freehd lfs_dlfs.dlfs_freehd 734 #define lfs_bfree lfs_dlfs.dlfs_bfree 735 #define lfs_nfiles lfs_dlfs.dlfs_nfiles 736 #define lfs_avail lfs_dlfs.dlfs_avail 737 #define lfs_uinodes lfs_dlfs.dlfs_uinodes 738 #define lfs_idaddr lfs_dlfs.dlfs_idaddr 739 #define lfs_ifile lfs_dlfs.dlfs_ifile 740 #define lfs_lastseg lfs_dlfs.dlfs_lastseg 741 #define lfs_nextseg lfs_dlfs.dlfs_nextseg 742 #define lfs_curseg lfs_dlfs.dlfs_curseg 743 #define lfs_offset lfs_dlfs.dlfs_offset 744 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg 745 #define lfs_otstamp lfs_dlfs.dlfs_inopf 746 #define lfs_inopf lfs_dlfs.dlfs_inopf 747 #define lfs_minfree lfs_dlfs.dlfs_minfree 748 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize 749 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg 750 #define lfs_inopb lfs_dlfs.dlfs_inopb 751 #define lfs_ifpb lfs_dlfs.dlfs_ifpb 752 #define lfs_sepb lfs_dlfs.dlfs_sepb 753 #define lfs_nindir lfs_dlfs.dlfs_nindir 754 #define lfs_nseg lfs_dlfs.dlfs_nseg 755 #define lfs_nspf lfs_dlfs.dlfs_nspf 756 #define lfs_cleansz lfs_dlfs.dlfs_cleansz 757 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz 758 #define lfs_segmask lfs_dlfs.dlfs_segmask 759 #define lfs_segshift lfs_dlfs.dlfs_segshift 760 #define lfs_bmask lfs_dlfs.dlfs_bmask 761 #define lfs_bshift lfs_dlfs.dlfs_bshift 762 #define lfs_ffmask lfs_dlfs.dlfs_ffmask 763 #define lfs_ffshift lfs_dlfs.dlfs_ffshift 764 #define lfs_fbmask lfs_dlfs.dlfs_fbmask 765 #define lfs_fbshift lfs_dlfs.dlfs_fbshift 766 #define lfs_blktodb lfs_dlfs.dlfs_blktodb 767 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb 768 #define lfs_sushift lfs_dlfs.dlfs_sushift 769 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen 770 #define lfs_sboffs lfs_dlfs.dlfs_sboffs 771 #define lfs_cksum lfs_dlfs.dlfs_cksum 772 #define lfs_pflags lfs_dlfs.dlfs_pflags 773 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt 774 #define lfs_nclean lfs_dlfs.dlfs_nclean 775 #define lfs_dmeta lfs_dlfs.dlfs_dmeta 776 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg 777 #define lfs_sumsize lfs_dlfs.dlfs_sumsize 778 #define lfs_serial lfs_dlfs.dlfs_serial 779 #define lfs_ibsize lfs_dlfs.dlfs_ibsize 780 #define lfs_start lfs_dlfs.dlfs_start 781 #define lfs_tstamp lfs_dlfs.dlfs_tstamp 782 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt 783 #define lfs_interleave lfs_dlfs.dlfs_interleave 784 #define lfs_ident lfs_dlfs.dlfs_ident 785 #define lfs_resvseg lfs_dlfs.dlfs_resvseg 786 787 /* These fields are set at mount time and are meaningless on disk. */ 788 struct segment *lfs_sp; /* current segment being written */ 789 struct vnode *lfs_ivnode; /* vnode for the ifile */ 790 u_int32_t lfs_seglock; /* single-thread the segment writer */ 791 pid_t lfs_lockpid; /* pid of lock holder */ 792 lwpid_t lfs_locklwp; /* lwp of lock holder */ 793 u_int32_t lfs_iocount; /* number of ios pending */ 794 u_int32_t lfs_writer; /* don't allow any dirops to start */ 795 u_int32_t lfs_dirops; /* count of active directory ops */ 796 u_int32_t lfs_dirvcount; /* count of VDIROP nodes in this fs */ 797 u_int32_t lfs_doifile; /* Write ifile blocks on next write */ 798 u_int32_t lfs_nactive; /* Number of segments since last ckp */ 799 int8_t lfs_fmod; /* super block modified flag */ 800 int8_t lfs_ronly; /* mounted read-only flag */ 801 #define LFS_NOTYET 0x01 802 #define LFS_IFDIRTY 0x02 803 #define LFS_WARNED 0x04 804 #define LFS_UNDIROP 0x08 805 int8_t lfs_flags; /* currently unused flag */ 806 u_int16_t lfs_activesb; /* toggle between superblocks */ 807 daddr_t lfs_sbactive; /* disk address of current sb write */ 808 struct vnode *lfs_flushvp; /* vnode being flushed */ 809 int lfs_flushvp_fakevref; /* fake vref count for flushvp */ 810 struct vnode *lfs_unlockvp; /* being inactivated in lfs_segunlock */ 811 u_int32_t lfs_diropwait; /* # procs waiting on dirop flush */ 812 size_t lfs_devbsize; /* Device block size */ 813 size_t lfs_devbshift; /* Device block shift */ 814 krwlock_t lfs_fraglock; 815 krwlock_t lfs_iflock; /* Ifile lock */ 816 kcondvar_t lfs_stopcv; /* Wrap lock */ 817 struct lwp *lfs_stoplwp; 818 pid_t lfs_rfpid; /* Process ID of roll-forward agent */ 819 int lfs_nadirop; /* number of active dirop nodes */ 820 long lfs_ravail; /* blocks pre-reserved for writing */ 821 long lfs_favail; /* blocks pre-reserved for writing */ 822 res_t *lfs_resblk; /* Reserved memory for pageout */ 823 TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */ 824 TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */ 825 #define LFS_RESHASH_WIDTH 17 826 LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH]; 827 int lfs_pdflush; /* pagedaemon wants us to flush */ 828 u_int32_t **lfs_suflags; /* Segment use flags */ 829 #ifdef _KERNEL 830 struct pool lfs_clpool; /* Pool for struct lfs_cluster */ 831 struct pool lfs_bpppool; /* Pool for bpp */ 832 struct pool lfs_segpool; /* Pool for struct segment */ 833 #endif /* _KERNEL */ 834 #define LFS_MAX_CLEANIND 64 835 int32_t lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */ 836 int lfs_cleanind; /* Index into intervals */ 837 int lfs_sleepers; /* # procs sleeping this fs */ 838 int lfs_pages; /* dirty pages blaming this fs */ 839 lfs_bm_t *lfs_ino_bitmap; /* Inuse inodes bitmap */ 840 int lfs_nowrap; /* Suspend log wrap */ 841 int lfs_wrappass; /* Allow first log wrap requester to pass */ 842 int lfs_wrapstatus; /* Wrap status */ 843 LIST_HEAD(, segdelta) lfs_segdhd; /* List of pending trunc accounting events */ 844 }; 845 846 /* NINDIR is the number of indirects in a file system block. */ 847 #define NINDIR(fs) ((fs)->lfs_nindir) 848 849 /* INOPB is the number of inodes in a secondary storage block. */ 850 #define INOPB(fs) ((fs)->lfs_inopb) 851 /* INOPF is the number of inodes in a fragment. */ 852 #define INOPF(fs) ((fs)->lfs_inopf) 853 854 #define blksize(fs, ip, lbn) \ 855 (((lbn) >= NDADDR || (ip)->i_ffs1_size >= ((lbn) + 1) << (fs)->lfs_bshift) \ 856 ? (fs)->lfs_bsize \ 857 : (fragroundup(fs, blkoff(fs, (ip)->i_ffs1_size)))) 858 #define blkoff(fs, loc) ((int)((loc) & (fs)->lfs_bmask)) 859 #define fragoff(fs, loc) /* calculates (loc % fs->lfs_fsize) */ \ 860 ((int)((loc) & (fs)->lfs_ffmask)) 861 862 #if defined (_KERNEL) 863 #define fsbtodb(fs, b) ((b) << ((fs)->lfs_ffshift - DEV_BSHIFT)) 864 #define dbtofsb(fs, b) ((b) >> ((fs)->lfs_ffshift - DEV_BSHIFT)) 865 #else 866 #define fsbtodb(fs, b) ((b) << (fs)->lfs_fsbtodb) 867 #define dbtofsb(fs, b) ((b) >> (fs)->lfs_fsbtodb) 868 #endif 869 870 #define lblkno(fs, loc) ((loc) >> (fs)->lfs_bshift) 871 #define lblktosize(fs, blk) ((blk) << (fs)->lfs_bshift) 872 873 #define fsbtob(fs, b) ((b) << (fs)->lfs_ffshift) 874 #define btofsb(fs, b) ((b) >> (fs)->lfs_ffshift) 875 876 #define numfrags(fs, loc) /* calculates (loc / fs->lfs_fsize) */ \ 877 ((loc) >> (fs)->lfs_ffshift) 878 #define blkroundup(fs, size) /* calculates roundup(size, fs->lfs_bsize) */ \ 879 ((off_t)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask))) 880 #define fragroundup(fs, size) /* calculates roundup(size, fs->lfs_fsize) */ \ 881 ((off_t)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask))) 882 #define fragstoblks(fs, frags)/* calculates (frags / fs->fs_frag) */ \ 883 ((frags) >> (fs)->lfs_fbshift) 884 #define blkstofrags(fs, blks) /* calculates (blks * fs->fs_frag) */ \ 885 ((blks) << (fs)->lfs_fbshift) 886 #define fragnum(fs, fsb) /* calculates (fsb % fs->lfs_frag) */ \ 887 ((fsb) & ((fs)->lfs_frag - 1)) 888 #define blknum(fs, fsb) /* calculates rounddown(fsb, fs->lfs_frag) */ \ 889 ((fsb) &~ ((fs)->lfs_frag - 1)) 890 #define dblksize(fs, dp, lbn) \ 891 (((lbn) >= NDADDR || (dp)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\ 892 ? (fs)->lfs_bsize \ 893 : (fragroundup(fs, blkoff(fs, (dp)->di_size)))) 894 895 #define segsize(fs) ((fs)->lfs_version == 1 ? \ 896 lblktosize((fs), (fs)->lfs_ssize) : \ 897 (fs)->lfs_ssize) 898 #define segtod(fs, seg) (((fs)->lfs_version == 1 ? \ 899 (fs)->lfs_ssize << (fs)->lfs_blktodb : \ 900 btofsb((fs), (fs)->lfs_ssize)) * (seg)) 901 #define dtosn(fs, daddr) /* block address to segment number */ \ 902 ((uint32_t)(((daddr) - (fs)->lfs_start) / segtod((fs), 1))) 903 #define sntod(fs, sn) /* segment number to disk address */ \ 904 ((daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start)) 905 906 /* 907 * Structures used by lfs_bmapv and lfs_markv to communicate information 908 * about inodes and data blocks. 909 */ 910 typedef struct block_info { 911 u_int32_t bi_inode; /* inode # */ 912 int32_t bi_lbn; /* logical block w/in file */ 913 int32_t bi_daddr; /* disk address of block */ 914 u_int64_t bi_segcreate; /* origin segment create time */ 915 int bi_version; /* file version number */ 916 void *bi_bp; /* data buffer */ 917 int bi_size; /* size of the block (if fragment) */ 918 } BLOCK_INFO; 919 920 /* Compatibility for 1.5 binaries */ 921 typedef struct block_info_15 { 922 u_int32_t bi_inode; /* inode # */ 923 int32_t bi_lbn; /* logical block w/in file */ 924 int32_t bi_daddr; /* disk address of block */ 925 u_int32_t bi_segcreate; /* origin segment create time */ 926 int bi_version; /* file version number */ 927 void *bi_bp; /* data buffer */ 928 int bi_size; /* size of the block (if fragment) */ 929 } BLOCK_INFO_15; 930 931 /* In-memory description of a segment about to be written. */ 932 struct segment { 933 struct lfs *fs; /* file system pointer */ 934 struct buf **bpp; /* pointer to buffer array */ 935 struct buf **cbpp; /* pointer to next available bp */ 936 struct buf **start_bpp; /* pointer to first bp in this set */ 937 struct buf *ibp; /* buffer pointer to inode page */ 938 struct ufs1_dinode *idp; /* pointer to ifile dinode */ 939 struct finfo *fip; /* current fileinfo pointer */ 940 struct vnode *vp; /* vnode being gathered */ 941 void *segsum; /* segment summary info */ 942 u_int32_t ninodes; /* number of inodes in this segment */ 943 int32_t seg_bytes_left; /* bytes left in segment */ 944 int32_t sum_bytes_left; /* bytes left in summary block */ 945 u_int32_t seg_number; /* number of this segment */ 946 int32_t *start_lbp; /* beginning lbn for this set */ 947 948 #define SEGM_CKP 0x01 /* doing a checkpoint */ 949 #define SEGM_CLEAN 0x02 /* cleaner call; don't sort */ 950 #define SEGM_SYNC 0x04 /* wait for segment */ 951 #define SEGM_PROT 0x08 /* don't inactivate at segunlock */ 952 #define SEGM_PAGEDAEMON 0x10 /* pagedaemon called us */ 953 #define SEGM_WRITERD 0x20 /* LFS writed called us */ 954 #define SEGM_FORCE_CKP 0x40 /* Force checkpoint right away */ 955 u_int16_t seg_flags; /* run-time flags for this segment */ 956 u_int32_t seg_iocount; /* number of ios pending */ 957 int ndupino; /* number of duplicate inodes */ 958 }; 959 960 #ifdef _KERNEL 961 struct lfs_cluster { 962 size_t bufsize; /* Size of kept data */ 963 struct buf **bpp; /* Array of kept buffers */ 964 int bufcount; /* Number of kept buffers */ 965 #define LFS_CL_MALLOC 0x00000001 966 #define LFS_CL_SHIFT 0x00000002 967 #define LFS_CL_SYNC 0x00000004 968 u_int32_t flags; /* Flags */ 969 struct lfs *fs; /* LFS that this belongs to */ 970 struct segment *seg; /* Segment structure, for LFS_CL_SYNC */ 971 }; 972 973 /* 974 * Splay tree containing block numbers allocated through lfs_balloc. 975 */ 976 struct lbnentry { 977 SPLAY_ENTRY(lbnentry) entry; 978 daddr_t lbn; 979 }; 980 #endif /* _KERNEL */ 981 982 /* 983 * LFS inode extensions. 984 */ 985 struct lfs_inode_ext { 986 off_t lfs_osize; /* size of file on disk */ 987 u_int32_t lfs_effnblocks; /* number of blocks when i/o completes */ 988 size_t lfs_fragsize[NDADDR]; /* size of on-disk direct blocks */ 989 TAILQ_ENTRY(inode) lfs_dchain; /* Dirop chain. */ 990 TAILQ_ENTRY(inode) lfs_pchain; /* Paging chain. */ 991 #define LFSI_NO_GOP_WRITE 0x01 992 #define LFSI_DELETED 0x02 993 #define LFSI_WRAPBLOCK 0x04 994 #define LFSI_WRAPWAIT 0x08 995 u_int32_t lfs_iflags; /* Inode flags */ 996 daddr_t lfs_hiblk; /* Highest lbn held by inode */ 997 #ifdef _KERNEL 998 SPLAY_HEAD(lfs_splay, lbnentry) lfs_lbtree; /* Tree of balloc'd lbns */ 999 int lfs_nbtree; /* Size of tree */ 1000 LIST_HEAD(, segdelta) lfs_segdhd; 1001 #endif 1002 int16_t lfs_odnlink; /* on-disk nlink count for cleaner */ 1003 }; 1004 #define i_lfs_osize inode_ext.lfs->lfs_osize 1005 #define i_lfs_effnblks inode_ext.lfs->lfs_effnblocks 1006 #define i_lfs_fragsize inode_ext.lfs->lfs_fragsize 1007 #define i_lfs_dchain inode_ext.lfs->lfs_dchain 1008 #define i_lfs_pchain inode_ext.lfs->lfs_pchain 1009 #define i_lfs_iflags inode_ext.lfs->lfs_iflags 1010 #define i_lfs_hiblk inode_ext.lfs->lfs_hiblk 1011 #define i_lfs_lbtree inode_ext.lfs->lfs_lbtree 1012 #define i_lfs_nbtree inode_ext.lfs->lfs_nbtree 1013 #define i_lfs_segdhd inode_ext.lfs->lfs_segdhd 1014 #define i_lfs_odnlink inode_ext.lfs->lfs_odnlink 1015 1016 /* 1017 * Macros for determining free space on the disk, with the variable metadata 1018 * of segment summaries and inode blocks taken into account. 1019 */ 1020 /* Estimate number of clean blocks not available for writing */ 1021 #define LFS_EST_CMETA(F) (int32_t)((((F)->lfs_dmeta * \ 1022 (int64_t)(F)->lfs_nclean) / \ 1023 ((F)->lfs_nseg - (F)->lfs_nclean))) 1024 1025 /* Estimate total size of the disk not including metadata */ 1026 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F)) 1027 1028 /* Estimate number of blocks actually available for writing */ 1029 #define LFS_EST_BFREE(F) ((F)->lfs_bfree > LFS_EST_CMETA(F) ? \ 1030 (F)->lfs_bfree - LFS_EST_CMETA(F) : 0) 1031 1032 /* Amount of non-meta space not available to mortal man */ 1033 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) * \ 1034 (u_int64_t)(F)->lfs_minfree) / \ 1035 100) 1036 1037 /* Can credential C write BB blocks */ 1038 #define ISSPACE(F, BB, C) \ 1039 ((((C) == NOCRED || kauth_cred_geteuid(C) == 0) && \ 1040 LFS_EST_BFREE(F) >= (BB)) || \ 1041 (kauth_cred_geteuid(C) != 0 && IS_FREESPACE(F, BB))) 1042 1043 /* Can an ordinary user write BB blocks */ 1044 #define IS_FREESPACE(F, BB) \ 1045 (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F)) 1046 1047 /* 1048 * The minimum number of blocks to create a new inode. This is: 1049 * directory direct block (1) + NIADDR indirect blocks + inode block (1) + 1050 * ifile direct block (1) + NIADDR indirect blocks = 3 + 2 * NIADDR blocks. 1051 */ 1052 #define LFS_NRESERVE(F) (btofsb((F), (2 * NIADDR + 3) << (F)->lfs_bshift)) 1053 1054 /* Statistics Counters */ 1055 struct lfs_stats { /* Must match sysctl list in lfs_vfsops.h ! */ 1056 u_int segsused; 1057 u_int psegwrites; 1058 u_int psyncwrites; 1059 u_int pcleanwrites; 1060 u_int blocktot; 1061 u_int cleanblocks; 1062 u_int ncheckpoints; 1063 u_int nwrites; 1064 u_int nsync_writes; 1065 u_int wait_exceeded; 1066 u_int write_exceeded; 1067 u_int flush_invoked; 1068 u_int vflush_invoked; 1069 u_int clean_inlocked; 1070 u_int clean_vnlocked; 1071 u_int segs_reclaimed; 1072 }; 1073 #ifdef _KERNEL 1074 extern struct lfs_stats lfs_stats; 1075 #endif 1076 1077 /* Fcntls to take the place of the lfs syscalls */ 1078 struct lfs_fcntl_markv { 1079 BLOCK_INFO *blkiov; /* blocks to relocate */ 1080 int blkcnt; /* number of blocks */ 1081 }; 1082 1083 #define LFCNSEGWAITALL _FCNR_FSPRIV('L', 14, struct timeval) 1084 #define LFCNSEGWAIT _FCNR_FSPRIV('L', 15, struct timeval) 1085 #define LFCNBMAPV _FCNRW_FSPRIV('L', 2, struct lfs_fcntl_markv) 1086 #define LFCNMARKV _FCNRW_FSPRIV('L', 3, struct lfs_fcntl_markv) 1087 #define LFCNRECLAIM _FCNO_FSPRIV('L', 4) 1088 1089 struct lfs_fhandle { 1090 char space[28]; /* FHANDLE_SIZE_COMPAT (but used from userland too) */ 1091 }; 1092 #define LFCNREWIND _FCNR_FSPRIV('L', 6, int) 1093 #define LFCNINVAL _FCNR_FSPRIV('L', 7, int) 1094 #define LFCNRESIZE _FCNR_FSPRIV('L', 8, int) 1095 #define LFCNWRAPSTOP _FCNR_FSPRIV('L', 9, int) 1096 #define LFCNWRAPGO _FCNR_FSPRIV('L', 10, int) 1097 #define LFCNIFILEFH _FCNW_FSPRIV('L', 11, struct lfs_fhandle) 1098 #define LFCNWRAPPASS _FCNR_FSPRIV('L', 12, int) 1099 # define LFS_WRAP_GOING 0x0 1100 # define LFS_WRAP_WAITING 0x1 1101 #define LFCNWRAPSTATUS _FCNW_FSPRIV('L', 13, int) 1102 1103 /* 1104 * Compat. Defined for kernel only. Userland always uses 1105 * "the one true version". 1106 */ 1107 #ifdef _KERNEL 1108 #include <compat/sys/time_types.h> 1109 1110 #define LFCNSEGWAITALL_COMPAT _FCNW_FSPRIV('L', 0, struct timeval50) 1111 #define LFCNSEGWAIT_COMPAT _FCNW_FSPRIV('L', 1, struct timeval50) 1112 #define LFCNIFILEFH_COMPAT _FCNW_FSPRIV('L', 5, struct lfs_fhandle) 1113 #define LFCNIFILEFH_COMPAT2 _FCN_FSPRIV(F_FSOUT, 'L', 11, 32) 1114 #define LFCNWRAPSTOP_COMPAT _FCNO_FSPRIV('L', 9) 1115 #define LFCNWRAPGO_COMPAT _FCNO_FSPRIV('L', 10) 1116 #define LFCNSEGWAITALL_COMPAT_50 _FCNR_FSPRIV('L', 0, struct timeval50) 1117 #define LFCNSEGWAIT_COMPAT_50 _FCNR_FSPRIV('L', 1, struct timeval50) 1118 #endif 1119 1120 #ifdef _KERNEL 1121 /* XXX MP */ 1122 #define LFS_SEGLOCK_HELD(fs) \ 1123 ((fs)->lfs_seglock != 0 && \ 1124 (fs)->lfs_lockpid == curproc->p_pid && \ 1125 (fs)->lfs_locklwp == curlwp->l_lid) 1126 #endif /* _KERNEL */ 1127 1128 /* Debug segment lock */ 1129 #ifdef notyet 1130 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs)) 1131 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs)) 1132 # define ASSERT_DUNNO_SEGLOCK(fs) 1133 # define ASSERT_MAYBE_SEGLOCK(fs) 1134 #else /* !notyet */ 1135 # define ASSERT_DUNNO_SEGLOCK(fs) \ 1136 DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \ 1137 LFS_SEGLOCK_HELD(fs))) 1138 # define ASSERT_SEGLOCK(fs) do { \ 1139 if (!LFS_SEGLOCK_HELD(fs)) { \ 1140 DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \ 1141 } \ 1142 } while(0) 1143 # define ASSERT_NO_SEGLOCK(fs) do { \ 1144 if (LFS_SEGLOCK_HELD(fs)) { \ 1145 DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \ 1146 } \ 1147 } while(0) 1148 # define ASSERT_MAYBE_SEGLOCK(x) 1149 #endif /* !notyet */ 1150 1151 __BEGIN_DECLS 1152 void lfs_itimes(struct inode *, const struct timespec *, 1153 const struct timespec *, const struct timespec *); 1154 __END_DECLS 1155 1156 #endif /* !_UFS_LFS_LFS_H_ */ 1157