lfs.h revision 1.143
1 /* $NetBSD: lfs.h,v 1.143 2013/06/08 02:09:35 dholland Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Konrad E. Schroder <perseant (at) hhhh.org>. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 /*- 32 * Copyright (c) 1991, 1993 33 * The Regents of the University of California. All rights reserved. 34 * 35 * Redistribution and use in source and binary forms, with or without 36 * modification, are permitted provided that the following conditions 37 * are met: 38 * 1. Redistributions of source code must retain the above copyright 39 * notice, this list of conditions and the following disclaimer. 40 * 2. Redistributions in binary form must reproduce the above copyright 41 * notice, this list of conditions and the following disclaimer in the 42 * documentation and/or other materials provided with the distribution. 43 * 3. Neither the name of the University nor the names of its contributors 44 * may be used to endorse or promote products derived from this software 45 * without specific prior written permission. 46 * 47 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 48 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 49 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 50 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 51 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 52 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 53 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 54 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 55 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 56 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 57 * SUCH DAMAGE. 58 * 59 * @(#)lfs.h 8.9 (Berkeley) 5/8/95 60 */ 61 62 #ifndef _UFS_LFS_LFS_H_ 63 #define _UFS_LFS_LFS_H_ 64 65 #include <sys/rwlock.h> 66 #include <sys/mutex.h> 67 #include <sys/queue.h> 68 #include <sys/condvar.h> 69 #include <sys/mount.h> 70 #include <sys/pool.h> 71 72 /* 73 * Compile-time options for LFS. 74 */ 75 #define LFS_IFIND_RETRIES 16 76 #define LFS_LOGLENGTH 1024 /* size of debugging log */ 77 #define LFS_MAX_ACTIVE 10 /* Dirty segments before ckp forced */ 78 79 /* 80 * Fixed filesystem layout parameters 81 */ 82 #define LFS_LABELPAD 8192 /* LFS label size */ 83 #define LFS_SBPAD 8192 /* LFS superblock size */ 84 85 #define LFS_UNUSED_INUM 0 /* 0: out of band inode number */ 86 #define LFS_IFILE_INUM 1 /* 1: IFILE inode number */ 87 /* 2: Root inode number */ 88 #define LFS_LOSTFOUNDINO 3 /* 3: lost+found inode number */ 89 #define LFS_FIRST_INUM 4 /* 4: first free inode number */ 90 91 #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 #define ULFS_NXADDR 2 99 #define ULFS_NDADDR 12 /* Direct addresses in inode. */ 100 #define ULFS_NIADDR 3 /* Indirect addresses in inode. */ 101 102 /* 103 * Adjustable filesystem parameters 104 */ 105 #define MIN_FREE_SEGS 20 106 #define MIN_RESV_SEGS 15 107 #ifndef LFS_ATIME_IFILE 108 # define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */ 109 #endif 110 #define LFS_MARKV_MAXBLKCNT 65536 /* Max block count for lfs_markv() */ 111 112 /* Misc. definitions */ 113 #define BW_CLEAN 1 /* Flag for lfs_bwrite_ext() */ 114 #define PG_DELWRI PG_PAGER1 /* Local def for delayed pageout */ 115 116 /* Resource limits */ 117 #define LFS_MAX_RESOURCE(x, u) (((x) >> 2) - 10 * (u)) 118 #define LFS_WAIT_RESOURCE(x, u) (((x) >> 1) - ((x) >> 3) - 10 * (u)) 119 #define LFS_INVERSE_MAX_RESOURCE(x, u) (((x) + 10 * (u)) << 2) 120 #define LFS_MAX_BUFS LFS_MAX_RESOURCE(nbuf, 1) 121 #define LFS_WAIT_BUFS LFS_WAIT_RESOURCE(nbuf, 1) 122 #define LFS_INVERSE_MAX_BUFS(n) LFS_INVERSE_MAX_RESOURCE(n, 1) 123 #define LFS_MAX_BYTES LFS_MAX_RESOURCE(bufmem_lowater, PAGE_SIZE) 124 #define LFS_INVERSE_MAX_BYTES(n) LFS_INVERSE_MAX_RESOURCE(n, PAGE_SIZE) 125 #define LFS_WAIT_BYTES LFS_WAIT_RESOURCE(bufmem_lowater, PAGE_SIZE) 126 #define LFS_MAX_DIROP ((desiredvnodes >> 2) + (desiredvnodes >> 3)) 127 #define SIZEOF_DIROP(fs) (2 * ((fs)->lfs_bsize + LFS_DINODE1_SIZE)) 128 #define LFS_MAX_FSDIROP(fs) \ 129 ((fs)->lfs_nclean <= (fs)->lfs_resvseg ? 0 : \ 130 (((fs)->lfs_nclean - (fs)->lfs_resvseg) * (fs)->lfs_ssize) / \ 131 (2 * SIZEOF_DIROP(fs))) 132 #define LFS_MAX_PAGES lfs_max_pages() 133 #define LFS_WAIT_PAGES lfs_wait_pages() 134 #define LFS_BUFWAIT 2 /* How long to wait if over *_WAIT_* */ 135 136 #ifdef _KERNEL 137 int lfs_wait_pages(void); 138 int lfs_max_pages(void); 139 #endif /* _KERNEL */ 140 141 /* How starved can we be before we start holding back page writes */ 142 #define LFS_STARVED_FOR_SEGS(fs) ((fs)->lfs_nclean < (fs)->lfs_resvseg) 143 144 /* 145 * Reserved blocks for lfs_malloc 146 */ 147 148 /* Structure to keep reserved blocks */ 149 typedef struct lfs_res_blk { 150 void *p; 151 LIST_ENTRY(lfs_res_blk) res; 152 int size; 153 char inuse; 154 } res_t; 155 156 /* Types for lfs_newbuf and lfs_malloc */ 157 #define LFS_NB_UNKNOWN -1 158 #define LFS_NB_SUMMARY 0 159 #define LFS_NB_SBLOCK 1 160 #define LFS_NB_IBLOCK 2 161 #define LFS_NB_CLUSTER 3 162 #define LFS_NB_CLEAN 4 163 #define LFS_NB_BLKIOV 5 164 #define LFS_NB_COUNT 6 /* always last */ 165 166 /* Number of reserved memory blocks of each type */ 167 #define LFS_N_SUMMARIES 2 168 #define LFS_N_SBLOCKS 1 /* Always 1, to throttle superblock writes */ 169 #define LFS_N_IBLOCKS 16 /* In theory ssize/bsize; in practice around 2 */ 170 #define LFS_N_CLUSTERS 16 /* In theory ssize/MAXPHYS */ 171 #define LFS_N_CLEAN 0 172 #define LFS_N_BLKIOV 1 173 174 /* Total count of "large" (non-pool) types */ 175 #define LFS_N_TOTAL (LFS_N_SUMMARIES + LFS_N_SBLOCKS + LFS_N_IBLOCKS + \ 176 LFS_N_CLUSTERS + LFS_N_CLEAN + LFS_N_BLKIOV) 177 178 /* Counts for pool types */ 179 #define LFS_N_CL LFS_N_CLUSTERS 180 #define LFS_N_BPP 2 181 #define LFS_N_SEG 2 182 183 /* 184 * Directories 185 */ 186 187 /* 188 * Theoretically, directories can be more than 2Gb in length; however, in 189 * practice this seems unlikely. So, we define the type doff_t as a 32-bit 190 * quantity to keep down the cost of doing lookup on a 32-bit machine. 191 */ 192 #define doff_t int32_t 193 #define lfs_doff_t int32_t 194 #define MAXDIRSIZE (0x7fffffff) 195 196 /* 197 * Inodes 198 */ 199 200 /* 201 * A dinode contains all the meta-data associated with a ULFS file. 202 * This structure defines the on-disk format of a dinode. Since 203 * this structure describes an on-disk structure, all its fields 204 * are defined by types with precise widths. 205 */ 206 207 struct ulfs1_dinode { 208 u_int16_t di_mode; /* 0: IFMT, permissions; see below. */ 209 int16_t di_nlink; /* 2: File link count. */ 210 union { 211 u_int16_t oldids[2]; /* 4: Ffs: old user and group ids. */ 212 u_int32_t inumber; /* 4: Lfs: inode number. */ 213 } di_u; 214 u_int64_t di_size; /* 8: File byte count. */ 215 int32_t di_atime; /* 16: Last access time. */ 216 int32_t di_atimensec; /* 20: Last access time. */ 217 int32_t di_mtime; /* 24: Last modified time. */ 218 int32_t di_mtimensec; /* 28: Last modified time. */ 219 int32_t di_ctime; /* 32: Last inode change time. */ 220 int32_t di_ctimensec; /* 36: Last inode change time. */ 221 int32_t di_db[ULFS_NDADDR]; /* 40: Direct disk blocks. */ 222 int32_t di_ib[ULFS_NIADDR]; /* 88: Indirect disk blocks. */ 223 u_int32_t di_flags; /* 100: Status flags (chflags). */ 224 u_int32_t di_blocks; /* 104: Blocks actually held. */ 225 int32_t di_gen; /* 108: Generation number. */ 226 u_int32_t di_uid; /* 112: File owner. */ 227 u_int32_t di_gid; /* 116: File group. */ 228 u_int64_t di_modrev; /* 120: i_modrev for NFSv4 */ 229 }; 230 231 struct ulfs2_dinode { 232 u_int16_t di_mode; /* 0: IFMT, permissions; see below. */ 233 int16_t di_nlink; /* 2: File link count. */ 234 u_int32_t di_uid; /* 4: File owner. */ 235 u_int32_t di_gid; /* 8: File group. */ 236 u_int32_t di_blksize; /* 12: Inode blocksize. */ 237 u_int64_t di_size; /* 16: File byte count. */ 238 u_int64_t di_blocks; /* 24: Bytes actually held. */ 239 int64_t di_atime; /* 32: Last access time. */ 240 int64_t di_mtime; /* 40: Last modified time. */ 241 int64_t di_ctime; /* 48: Last inode change time. */ 242 int64_t di_birthtime; /* 56: Inode creation time. */ 243 int32_t di_mtimensec; /* 64: Last modified time. */ 244 int32_t di_atimensec; /* 68: Last access time. */ 245 int32_t di_ctimensec; /* 72: Last inode change time. */ 246 int32_t di_birthnsec; /* 76: Inode creation time. */ 247 int32_t di_gen; /* 80: Generation number. */ 248 u_int32_t di_kernflags; /* 84: Kernel flags. */ 249 u_int32_t di_flags; /* 88: Status flags (chflags). */ 250 int32_t di_extsize; /* 92: External attributes block. */ 251 int64_t di_extb[ULFS_NXADDR];/* 96: External attributes block. */ 252 int64_t di_db[ULFS_NDADDR]; /* 112: Direct disk blocks. */ 253 int64_t di_ib[ULFS_NIADDR]; /* 208: Indirect disk blocks. */ 254 u_int64_t di_modrev; /* 232: i_modrev for NFSv4 */ 255 int64_t di_spare[2]; /* 240: Reserved; currently unused */ 256 }; 257 258 /* 259 * The di_db fields may be overlaid with other information for 260 * file types that do not have associated disk storage. Block 261 * and character devices overlay the first data block with their 262 * dev_t value. Short symbolic links place their path in the 263 * di_db area. 264 */ 265 #define di_inumber di_u.inumber 266 #define di_ogid di_u.oldids[1] 267 #define di_ouid di_u.oldids[0] 268 #define di_rdev di_db[0] 269 270 /* Size of the on-disk inode. */ 271 #define LFS_DINODE1_SIZE (sizeof(struct ulfs1_dinode)) /* 128 */ 272 #define LFS_DINODE2_SIZE (sizeof(struct ulfs2_dinode)) 273 274 /* 275 * "struct buf" associated definitions 276 */ 277 278 /* Unassigned disk addresses. */ 279 #define UNASSIGNED -1 280 #define UNWRITTEN -2 281 282 /* Unused logical block number */ 283 #define LFS_UNUSED_LBN -1 284 285 /* Determine if a buffer belongs to the ifile */ 286 #define IS_IFILE(bp) (VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM) 287 288 # define LFS_LOCK_BUF(bp) do { \ 289 if (((bp)->b_flags & B_LOCKED) == 0 && bp->b_iodone == NULL) { \ 290 mutex_enter(&lfs_lock); \ 291 ++locked_queue_count; \ 292 locked_queue_bytes += bp->b_bufsize; \ 293 mutex_exit(&lfs_lock); \ 294 } \ 295 (bp)->b_flags |= B_LOCKED; \ 296 } while (0) 297 298 # define LFS_UNLOCK_BUF(bp) do { \ 299 if (((bp)->b_flags & B_LOCKED) != 0 && bp->b_iodone == NULL) { \ 300 mutex_enter(&lfs_lock); \ 301 --locked_queue_count; \ 302 locked_queue_bytes -= bp->b_bufsize; \ 303 if (locked_queue_count < LFS_WAIT_BUFS && \ 304 locked_queue_bytes < LFS_WAIT_BYTES) \ 305 cv_broadcast(&locked_queue_cv); \ 306 mutex_exit(&lfs_lock); \ 307 } \ 308 (bp)->b_flags &= ~B_LOCKED; \ 309 } while (0) 310 311 #ifdef _KERNEL 312 313 extern u_long bufmem_lowater, bufmem_hiwater; /* XXX */ 314 315 # define LFS_IS_MALLOC_BUF(bp) ((bp)->b_iodone == lfs_callback) 316 317 # ifdef DEBUG 318 # define LFS_DEBUG_COUNTLOCKED(m) do { \ 319 if (lfs_debug_log_subsys[DLOG_LLIST]) { \ 320 lfs_countlocked(&locked_queue_count, &locked_queue_bytes, (m)); \ 321 cv_broadcast(&locked_queue_cv); \ 322 } \ 323 } while (0) 324 # else 325 # define LFS_DEBUG_COUNTLOCKED(m) 326 # endif 327 328 /* log for debugging writes to the Ifile */ 329 # ifdef DEBUG 330 struct lfs_log_entry { 331 const char *op; 332 const char *file; 333 int pid; 334 int line; 335 daddr_t block; 336 unsigned long flags; 337 }; 338 extern int lfs_lognum; 339 extern struct lfs_log_entry lfs_log[LFS_LOGLENGTH]; 340 # define LFS_BWRITE_LOG(bp) lfs_bwrite_log((bp), __FILE__, __LINE__) 341 # define LFS_ENTER_LOG(theop, thefile, theline, lbn, theflags, thepid) do {\ 342 int _s; \ 343 \ 344 mutex_enter(&lfs_lock); \ 345 _s = splbio(); \ 346 lfs_log[lfs_lognum].op = theop; \ 347 lfs_log[lfs_lognum].file = thefile; \ 348 lfs_log[lfs_lognum].line = (theline); \ 349 lfs_log[lfs_lognum].pid = (thepid); \ 350 lfs_log[lfs_lognum].block = (lbn); \ 351 lfs_log[lfs_lognum].flags = (theflags); \ 352 lfs_lognum = (lfs_lognum + 1) % LFS_LOGLENGTH; \ 353 splx(_s); \ 354 mutex_exit(&lfs_lock); \ 355 } while (0) 356 357 # define LFS_BCLEAN_LOG(fs, bp) do { \ 358 if ((bp)->b_vp == (fs)->lfs_ivnode) \ 359 LFS_ENTER_LOG("clear", __FILE__, __LINE__, \ 360 bp->b_lblkno, bp->b_flags, curproc->p_pid);\ 361 } while (0) 362 363 /* Must match list in lfs_vfsops.c ! */ 364 # define DLOG_RF 0 /* roll forward */ 365 # define DLOG_ALLOC 1 /* inode alloc */ 366 # define DLOG_AVAIL 2 /* lfs_{,r,f}avail */ 367 # define DLOG_FLUSH 3 /* flush */ 368 # define DLOG_LLIST 4 /* locked list accounting */ 369 # define DLOG_WVNODE 5 /* vflush/writevnodes verbose */ 370 # define DLOG_VNODE 6 /* vflush/writevnodes */ 371 # define DLOG_SEG 7 /* segwrite */ 372 # define DLOG_SU 8 /* seguse accounting */ 373 # define DLOG_CLEAN 9 /* cleaner routines */ 374 # define DLOG_MOUNT 10 /* mount/unmount */ 375 # define DLOG_PAGE 11 /* putpages/gop_write */ 376 # define DLOG_DIROP 12 /* dirop accounting */ 377 # define DLOG_MALLOC 13 /* lfs_malloc accounting */ 378 # define DLOG_MAX 14 /* The terminator */ 379 # define DLOG(a) lfs_debug_log a 380 # else /* ! DEBUG */ 381 # define LFS_BCLEAN_LOG(fs, bp) 382 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)->b_vp, (bp)) 383 # define DLOG(a) 384 # endif /* ! DEBUG */ 385 #else /* ! _KERNEL */ 386 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)) 387 #endif /* _KERNEL */ 388 389 #ifdef _KERNEL 390 /* This overlays the fid structure (see fstypes.h). */ 391 struct ulfs_ufid { 392 u_int16_t ufid_len; /* Length of structure. */ 393 u_int16_t ufid_pad; /* Force 32-bit alignment. */ 394 u_int32_t ufid_ino; /* File number (ino). */ 395 int32_t ufid_gen; /* Generation number. */ 396 }; 397 /* Filehandle structure for exported LFSes */ 398 struct lfid { 399 struct ulfs_ufid lfid_ufid; 400 #define lfid_len lfid_ufid.ufid_len 401 #define lfid_ino lfid_ufid.ufid_ino 402 #define lfid_gen lfid_ufid.ufid_gen 403 uint32_t lfid_ident; 404 }; 405 #endif /* _KERNEL */ 406 407 /* 408 * "struct inode" associated definitions 409 */ 410 411 /* Address calculations for metadata located in the inode */ 412 #define S_INDIR(fs) -ULFS_NDADDR 413 #define D_INDIR(fs) (S_INDIR(fs) - NINDIR(fs) - 1) 414 #define T_INDIR(fs) (D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1) 415 416 /* For convenience */ 417 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING) 418 419 #define LFS_SET_UINO(ip, flags) do { \ 420 if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED)) \ 421 ++(ip)->i_lfs->lfs_uinodes; \ 422 if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING)) \ 423 ++(ip)->i_lfs->lfs_uinodes; \ 424 if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED)) \ 425 ++(ip)->i_lfs->lfs_uinodes; \ 426 (ip)->i_flag |= (flags); \ 427 } while (0) 428 429 #define LFS_CLR_UINO(ip, flags) do { \ 430 if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED)) \ 431 --(ip)->i_lfs->lfs_uinodes; \ 432 if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING)) \ 433 --(ip)->i_lfs->lfs_uinodes; \ 434 if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED)) \ 435 --(ip)->i_lfs->lfs_uinodes; \ 436 (ip)->i_flag &= ~(flags); \ 437 if ((ip)->i_lfs->lfs_uinodes < 0) { \ 438 panic("lfs_uinodes < 0"); \ 439 } \ 440 } while (0) 441 442 #define LFS_ITIMES(ip, acc, mod, cre) \ 443 while ((ip)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY)) \ 444 lfs_itimes(ip, acc, mod, cre) 445 446 /* 447 * "struct vnode" associated definitions 448 */ 449 450 /* Heuristic emptiness measure */ 451 #define VPISEMPTY(vp) (LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 452 !(vp->v_type == VREG && (vp)->v_iflag & VI_ONWORKLST) &&\ 453 VTOI(vp)->i_lfs_nbtree == 0) 454 455 #define WRITEINPROG(vp) ((vp)->v_numoutput > 0 || \ 456 (!LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 457 !(VTOI(vp)->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING)))) 458 459 460 /* 461 * On-disk and in-memory checkpoint segment usage structure. 462 */ 463 typedef struct segusage SEGUSE; 464 struct segusage { 465 u_int32_t su_nbytes; /* 0: number of live bytes */ 466 u_int32_t su_olastmod; /* 4: SEGUSE last modified timestamp */ 467 u_int16_t su_nsums; /* 8: number of summaries in segment */ 468 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 469 470 #define SEGUSE_ACTIVE 0x01 /* segment currently being written */ 471 #define SEGUSE_DIRTY 0x02 /* segment has data in it */ 472 #define SEGUSE_SUPERBLOCK 0x04 /* segment contains a superblock */ 473 #define SEGUSE_ERROR 0x08 /* cleaner: do not clean segment */ 474 #define SEGUSE_EMPTY 0x10 /* segment is empty */ 475 #define SEGUSE_INVAL 0x20 /* segment is invalid */ 476 u_int32_t su_flags; /* 12: segment flags */ 477 u_int64_t su_lastmod; /* 16: last modified timestamp */ 478 }; 479 480 typedef struct segusage_v1 SEGUSE_V1; 481 struct segusage_v1 { 482 u_int32_t su_nbytes; /* 0: number of live bytes */ 483 u_int32_t su_lastmod; /* 4: SEGUSE last modified timestamp */ 484 u_int16_t su_nsums; /* 8: number of summaries in segment */ 485 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 486 u_int32_t su_flags; /* 12: segment flags */ 487 }; 488 489 #define SEGUPB(fs) (fs->lfs_sepb) 490 #define SEGTABSIZE_SU(fs) \ 491 (((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb) 492 493 #ifdef _KERNEL 494 # define SHARE_IFLOCK(F) \ 495 do { \ 496 rw_enter(&(F)->lfs_iflock, RW_READER); \ 497 } while(0) 498 # define UNSHARE_IFLOCK(F) \ 499 do { \ 500 rw_exit(&(F)->lfs_iflock); \ 501 } while(0) 502 #else /* ! _KERNEL */ 503 # define SHARE_IFLOCK(F) 504 # define UNSHARE_IFLOCK(F) 505 #endif /* ! _KERNEL */ 506 507 /* Read in the block with a specific segment usage entry from the ifile. */ 508 #define LFS_SEGENTRY(SP, F, IN, BP) do { \ 509 int _e; \ 510 SHARE_IFLOCK(F); \ 511 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 512 if ((_e = bread((F)->lfs_ivnode, \ 513 ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz, \ 514 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 515 panic("lfs: ifile read: %d", _e); \ 516 if ((F)->lfs_version == 1) \ 517 (SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data + \ 518 ((IN) & ((F)->lfs_sepb - 1))); \ 519 else \ 520 (SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb); \ 521 UNSHARE_IFLOCK(F); \ 522 } while (0) 523 524 #define LFS_WRITESEGENTRY(SP, F, IN, BP) do { \ 525 if ((SP)->su_nbytes == 0) \ 526 (SP)->su_flags |= SEGUSE_EMPTY; \ 527 else \ 528 (SP)->su_flags &= ~SEGUSE_EMPTY; \ 529 (F)->lfs_suflags[(F)->lfs_activesb][(IN)] = (SP)->su_flags; \ 530 LFS_BWRITE_LOG(BP); \ 531 } while (0) 532 533 /* 534 * On-disk file information. One per file with data blocks in the segment. 535 */ 536 typedef struct finfo FINFO; 537 struct finfo { 538 u_int32_t fi_nblocks; /* number of blocks */ 539 u_int32_t fi_version; /* version number */ 540 u_int32_t fi_ino; /* inode number */ 541 u_int32_t fi_lastlength; /* length of last block in array */ 542 int32_t fi_blocks[1]; /* array of logical block numbers */ 543 }; 544 /* sizeof FINFO except fi_blocks */ 545 #define FINFOSIZE (sizeof(FINFO) - sizeof(int32_t)) 546 547 /* 548 * Index file inode entries. 549 */ 550 typedef struct ifile IFILE; 551 struct ifile { 552 u_int32_t if_version; /* inode version number */ 553 #define LFS_UNUSED_DADDR 0 /* out-of-band daddr */ 554 int32_t if_daddr; /* inode disk address */ 555 #define LFS_ORPHAN_NEXTFREE (~(u_int32_t)0) /* indicate orphaned file */ 556 u_int32_t if_nextfree; /* next-unallocated inode */ 557 u_int32_t if_atime_sec; /* Last access time, seconds */ 558 u_int32_t if_atime_nsec; /* and nanoseconds */ 559 }; 560 561 typedef struct ifile_v1 IFILE_V1; 562 struct ifile_v1 { 563 u_int32_t if_version; /* inode version number */ 564 int32_t if_daddr; /* inode disk address */ 565 u_int32_t if_nextfree; /* next-unallocated inode */ 566 #if LFS_ATIME_IFILE 567 struct timespec if_atime; /* Last access time */ 568 #endif 569 }; 570 571 /* 572 * LFSv1 compatibility code is not allowed to touch if_atime, since it 573 * may not be mapped! 574 */ 575 /* Read in the block with a specific inode from the ifile. */ 576 #define LFS_IENTRY(IP, F, IN, BP) do { \ 577 int _e; \ 578 SHARE_IFLOCK(F); \ 579 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 580 if ((_e = bread((F)->lfs_ivnode, \ 581 (IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz, \ 582 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 583 panic("lfs: ifile ino %d read %d", (int)(IN), _e); \ 584 if ((F)->lfs_version == 1) \ 585 (IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data + \ 586 (IN) % (F)->lfs_ifpb); \ 587 else \ 588 (IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb; \ 589 UNSHARE_IFLOCK(F); \ 590 } while (0) 591 592 /* 593 * Cleaner information structure. This resides in the ifile and is used 594 * to pass information from the kernel to the cleaner. 595 */ 596 typedef struct _cleanerinfo { 597 u_int32_t clean; /* number of clean segments */ 598 u_int32_t dirty; /* number of dirty segments */ 599 int32_t bfree; /* disk blocks free */ 600 int32_t avail; /* disk blocks available */ 601 u_int32_t free_head; /* head of the inode free list */ 602 u_int32_t free_tail; /* tail of the inode free list */ 603 #define LFS_CLEANER_MUST_CLEAN 0x01 604 u_int32_t flags; /* status word from the kernel */ 605 } CLEANERINFO; 606 607 #define CLEANSIZE_SU(fs) \ 608 ((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift) 609 610 /* Read in the block with the cleaner info from the ifile. */ 611 #define LFS_CLEANERINFO(CP, F, BP) do { \ 612 SHARE_IFLOCK(F); \ 613 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 614 if (bread((F)->lfs_ivnode, \ 615 (daddr_t)0, (F)->lfs_bsize, NOCRED, 0, &(BP))) \ 616 panic("lfs: ifile read"); \ 617 (CP) = (CLEANERINFO *)(BP)->b_data; \ 618 UNSHARE_IFLOCK(F); \ 619 } while (0) 620 621 /* 622 * Synchronize the Ifile cleaner info with current avail and bfree. 623 */ 624 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do { \ 625 mutex_enter(&lfs_lock); \ 626 if ((w) || (cip)->bfree != (fs)->lfs_bfree || \ 627 (cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail - \ 628 (fs)->lfs_favail) { \ 629 (cip)->bfree = (fs)->lfs_bfree; \ 630 (cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail - \ 631 (fs)->lfs_favail; \ 632 if (((bp)->b_flags & B_GATHERED) == 0) { \ 633 (fs)->lfs_flags |= LFS_IFDIRTY; \ 634 } \ 635 mutex_exit(&lfs_lock); \ 636 (void) LFS_BWRITE_LOG(bp); /* Ifile */ \ 637 } else { \ 638 mutex_exit(&lfs_lock); \ 639 brelse(bp, 0); \ 640 } \ 641 } while (0) 642 643 /* 644 * Get the head of the inode free list. 645 * Always called with the segment lock held. 646 */ 647 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do { \ 648 if ((FS)->lfs_version > 1) { \ 649 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 650 (FS)->lfs_freehd = (CIP)->free_head; \ 651 brelse(BP, 0); \ 652 } \ 653 *(FREEP) = (FS)->lfs_freehd; \ 654 } while (0) 655 656 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do { \ 657 (FS)->lfs_freehd = (VAL); \ 658 if ((FS)->lfs_version > 1) { \ 659 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 660 (CIP)->free_head = (VAL); \ 661 LFS_BWRITE_LOG(BP); \ 662 mutex_enter(&lfs_lock); \ 663 (FS)->lfs_flags |= LFS_IFDIRTY; \ 664 mutex_exit(&lfs_lock); \ 665 } \ 666 } while (0) 667 668 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do { \ 669 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 670 *(FREEP) = (CIP)->free_tail; \ 671 brelse(BP, 0); \ 672 } while (0) 673 674 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do { \ 675 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 676 (CIP)->free_tail = (VAL); \ 677 LFS_BWRITE_LOG(BP); \ 678 mutex_enter(&lfs_lock); \ 679 (FS)->lfs_flags |= LFS_IFDIRTY; \ 680 mutex_exit(&lfs_lock); \ 681 } while (0) 682 683 /* 684 * On-disk segment summary information 685 */ 686 typedef struct segsum_v1 SEGSUM_V1; 687 struct segsum_v1 { 688 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 689 u_int32_t ss_datasum; /* 4: check sum of data */ 690 u_int32_t ss_magic; /* 8: segment summary magic number */ 691 #define SS_MAGIC 0x061561 692 int32_t ss_next; /* 12: next segment */ 693 u_int32_t ss_create; /* 16: creation time stamp */ 694 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 695 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 696 697 #define SS_DIROP 0x01 /* segment begins a dirop */ 698 #define SS_CONT 0x02 /* more partials to finish this write*/ 699 #define SS_CLEAN 0x04 /* written by the cleaner */ 700 #define SS_RFW 0x08 /* written by the roll-forward agent */ 701 #define SS_RECLAIM 0x10 /* written by the roll-forward agent */ 702 u_int16_t ss_flags; /* 24: used for directory operations */ 703 u_int16_t ss_pad; /* 26: extra space */ 704 /* FINFO's and inode daddr's... */ 705 }; 706 707 typedef struct segsum SEGSUM; 708 struct segsum { 709 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 710 u_int32_t ss_datasum; /* 4: check sum of data */ 711 u_int32_t ss_magic; /* 8: segment summary magic number */ 712 int32_t ss_next; /* 12: next segment */ 713 u_int32_t ss_ident; /* 16: roll-forward fsid */ 714 #define ss_ocreate ss_ident /* ident is where create was in v1 */ 715 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 716 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 717 u_int16_t ss_flags; /* 24: used for directory operations */ 718 u_int8_t ss_pad[2]; /* 26: extra space */ 719 u_int32_t ss_reclino; /* 28: inode being reclaimed */ 720 u_int64_t ss_serial; /* 32: serial number */ 721 u_int64_t ss_create; /* 40: time stamp */ 722 /* FINFO's and inode daddr's... */ 723 }; 724 725 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM)) 726 727 728 /* 729 * On-disk super block. 730 */ 731 struct dlfs { 732 #define LFS_MAGIC 0x070162 733 u_int32_t dlfs_magic; /* 0: magic number */ 734 #define LFS_VERSION 2 735 u_int32_t dlfs_version; /* 4: version number */ 736 737 u_int32_t dlfs_size; /* 8: number of blocks in fs (v1) */ 738 /* number of frags in fs (v2) */ 739 u_int32_t dlfs_ssize; /* 12: number of blocks per segment (v1) */ 740 /* number of bytes per segment (v2) */ 741 u_int32_t dlfs_dsize; /* 16: number of disk blocks in fs */ 742 u_int32_t dlfs_bsize; /* 20: file system block size */ 743 u_int32_t dlfs_fsize; /* 24: size of frag blocks in fs */ 744 u_int32_t dlfs_frag; /* 28: number of frags in a block in fs */ 745 746 /* Checkpoint region. */ 747 u_int32_t dlfs_freehd; /* 32: start of the free list */ 748 int32_t dlfs_bfree; /* 36: number of free disk blocks */ 749 u_int32_t dlfs_nfiles; /* 40: number of allocated inodes */ 750 int32_t dlfs_avail; /* 44: blocks available for writing */ 751 int32_t dlfs_uinodes; /* 48: inodes in cache not yet on disk */ 752 int32_t dlfs_idaddr; /* 52: inode file disk address */ 753 u_int32_t dlfs_ifile; /* 56: inode file inode number */ 754 int32_t dlfs_lastseg; /* 60: address of last segment written */ 755 int32_t dlfs_nextseg; /* 64: address of next segment to write */ 756 int32_t dlfs_curseg; /* 68: current segment being written */ 757 int32_t dlfs_offset; /* 72: offset in curseg for next partial */ 758 int32_t dlfs_lastpseg; /* 76: address of last partial written */ 759 u_int32_t dlfs_inopf; /* 80: v1: time stamp; v2: inodes per frag */ 760 #define dlfs_otstamp dlfs_inopf 761 762 /* These are configuration parameters. */ 763 u_int32_t dlfs_minfree; /* 84: minimum percentage of free blocks */ 764 765 /* These fields can be computed from the others. */ 766 u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */ 767 u_int32_t dlfs_fsbpseg; /* 96: fsb per segment */ 768 u_int32_t dlfs_inopb; /* 100: inodes per block */ 769 u_int32_t dlfs_ifpb; /* 104: IFILE entries per block */ 770 u_int32_t dlfs_sepb; /* 108: SEGUSE entries per block */ 771 u_int32_t dlfs_nindir; /* 112: indirect pointers per block */ 772 u_int32_t dlfs_nseg; /* 116: number of segments */ 773 u_int32_t dlfs_nspf; /* 120: number of sectors per fragment */ 774 u_int32_t dlfs_cleansz; /* 124: cleaner info size in blocks */ 775 u_int32_t dlfs_segtabsz; /* 128: segment table size in blocks */ 776 u_int32_t dlfs_segmask; /* 132: calculate offset within a segment */ 777 u_int32_t dlfs_segshift; /* 136: fast mult/div for segments */ 778 u_int32_t dlfs_bshift; /* 140: calc block number from file offset */ 779 u_int32_t dlfs_ffshift; /* 144: fast mult/div for frag from file */ 780 u_int32_t dlfs_fbshift; /* 148: fast mult/div for frag from block */ 781 u_int64_t dlfs_bmask; /* 152: calc block offset from file offset */ 782 u_int64_t dlfs_ffmask; /* 160: calc frag offset from file offset */ 783 u_int64_t dlfs_fbmask; /* 168: calc frag offset from block offset */ 784 u_int32_t dlfs_blktodb; /* 176: blktodb and dbtoblk shift constant */ 785 u_int32_t dlfs_sushift; /* 180: fast mult/div for segusage table */ 786 787 int32_t dlfs_maxsymlinklen; /* 184: max length of an internal symlink */ 788 #define LFS_MIN_SBINTERVAL 5 /* minimum superblock segment spacing */ 789 #define LFS_MAXNUMSB 10 /* 188: superblock disk offsets */ 790 int32_t dlfs_sboffs[LFS_MAXNUMSB]; 791 792 u_int32_t dlfs_nclean; /* 228: Number of clean segments */ 793 u_char dlfs_fsmnt[MNAMELEN]; /* 232: name mounted on */ 794 #define LFS_PF_CLEAN 0x1 795 u_int16_t dlfs_pflags; /* 322: file system persistent flags */ 796 int32_t dlfs_dmeta; /* 324: total number of dirty summaries */ 797 u_int32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */ 798 u_int32_t dlfs_sumsize; /* 332: size of summary blocks */ 799 u_int64_t dlfs_serial; /* 336: serial number */ 800 u_int32_t dlfs_ibsize; /* 344: size of inode blocks */ 801 int32_t dlfs_start; /* 348: start of segment 0 */ 802 u_int64_t dlfs_tstamp; /* 352: time stamp */ 803 #define LFS_44INODEFMT 0 804 #define LFS_MAXINODEFMT 0 805 u_int32_t dlfs_inodefmt; /* 360: inode format version */ 806 u_int32_t dlfs_interleave; /* 364: segment interleave */ 807 u_int32_t dlfs_ident; /* 368: per-fs identifier */ 808 u_int32_t dlfs_fsbtodb; /* 372: fsbtodb abd dbtodsb shift constant */ 809 u_int32_t dlfs_resvseg; /* 376: segments reserved for the cleaner */ 810 int8_t dlfs_pad[128]; /* 380: round to 512 bytes */ 811 /* Checksum -- last valid disk field. */ 812 u_int32_t dlfs_cksum; /* 508: checksum for superblock checking */ 813 }; 814 815 /* Type used for the inode bitmap */ 816 typedef u_int32_t lfs_bm_t; 817 818 /* 819 * Linked list of segments whose byte count needs updating following a 820 * file truncation. 821 */ 822 struct segdelta { 823 long segnum; 824 size_t num; 825 LIST_ENTRY(segdelta) list; 826 }; 827 828 /* 829 * In-memory super block. 830 */ 831 struct lfs { 832 struct dlfs lfs_dlfs; /* on-disk parameters */ 833 #define lfs_magic lfs_dlfs.dlfs_magic 834 #define lfs_version lfs_dlfs.dlfs_version 835 #define lfs_size lfs_dlfs.dlfs_size 836 #define lfs_ssize lfs_dlfs.dlfs_ssize 837 #define lfs_dsize lfs_dlfs.dlfs_dsize 838 #define lfs_bsize lfs_dlfs.dlfs_bsize 839 #define lfs_fsize lfs_dlfs.dlfs_fsize 840 #define lfs_frag lfs_dlfs.dlfs_frag 841 #define lfs_freehd lfs_dlfs.dlfs_freehd 842 #define lfs_bfree lfs_dlfs.dlfs_bfree 843 #define lfs_nfiles lfs_dlfs.dlfs_nfiles 844 #define lfs_avail lfs_dlfs.dlfs_avail 845 #define lfs_uinodes lfs_dlfs.dlfs_uinodes 846 #define lfs_idaddr lfs_dlfs.dlfs_idaddr 847 #define lfs_ifile lfs_dlfs.dlfs_ifile 848 #define lfs_lastseg lfs_dlfs.dlfs_lastseg 849 #define lfs_nextseg lfs_dlfs.dlfs_nextseg 850 #define lfs_curseg lfs_dlfs.dlfs_curseg 851 #define lfs_offset lfs_dlfs.dlfs_offset 852 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg 853 #define lfs_otstamp lfs_dlfs.dlfs_inopf 854 #define lfs_inopf lfs_dlfs.dlfs_inopf 855 #define lfs_minfree lfs_dlfs.dlfs_minfree 856 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize 857 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg 858 #define lfs_inopb lfs_dlfs.dlfs_inopb 859 #define lfs_ifpb lfs_dlfs.dlfs_ifpb 860 #define lfs_sepb lfs_dlfs.dlfs_sepb 861 #define lfs_nindir lfs_dlfs.dlfs_nindir 862 #define lfs_nseg lfs_dlfs.dlfs_nseg 863 #define lfs_nspf lfs_dlfs.dlfs_nspf 864 #define lfs_cleansz lfs_dlfs.dlfs_cleansz 865 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz 866 #define lfs_segmask lfs_dlfs.dlfs_segmask 867 #define lfs_segshift lfs_dlfs.dlfs_segshift 868 #define lfs_bmask lfs_dlfs.dlfs_bmask 869 #define lfs_bshift lfs_dlfs.dlfs_bshift 870 #define lfs_ffmask lfs_dlfs.dlfs_ffmask 871 #define lfs_ffshift lfs_dlfs.dlfs_ffshift 872 #define lfs_fbmask lfs_dlfs.dlfs_fbmask 873 #define lfs_fbshift lfs_dlfs.dlfs_fbshift 874 #define lfs_blktodb lfs_dlfs.dlfs_blktodb 875 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb 876 #define lfs_sushift lfs_dlfs.dlfs_sushift 877 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen 878 #define lfs_sboffs lfs_dlfs.dlfs_sboffs 879 #define lfs_cksum lfs_dlfs.dlfs_cksum 880 #define lfs_pflags lfs_dlfs.dlfs_pflags 881 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt 882 #define lfs_nclean lfs_dlfs.dlfs_nclean 883 #define lfs_dmeta lfs_dlfs.dlfs_dmeta 884 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg 885 #define lfs_sumsize lfs_dlfs.dlfs_sumsize 886 #define lfs_serial lfs_dlfs.dlfs_serial 887 #define lfs_ibsize lfs_dlfs.dlfs_ibsize 888 #define lfs_start lfs_dlfs.dlfs_start 889 #define lfs_tstamp lfs_dlfs.dlfs_tstamp 890 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt 891 #define lfs_interleave lfs_dlfs.dlfs_interleave 892 #define lfs_ident lfs_dlfs.dlfs_ident 893 #define lfs_resvseg lfs_dlfs.dlfs_resvseg 894 895 /* These fields are set at mount time and are meaningless on disk. */ 896 struct segment *lfs_sp; /* current segment being written */ 897 struct vnode *lfs_ivnode; /* vnode for the ifile */ 898 u_int32_t lfs_seglock; /* single-thread the segment writer */ 899 pid_t lfs_lockpid; /* pid of lock holder */ 900 lwpid_t lfs_locklwp; /* lwp of lock holder */ 901 u_int32_t lfs_iocount; /* number of ios pending */ 902 u_int32_t lfs_writer; /* don't allow any dirops to start */ 903 u_int32_t lfs_dirops; /* count of active directory ops */ 904 u_int32_t lfs_dirvcount; /* count of VDIROP nodes in this fs */ 905 u_int32_t lfs_doifile; /* Write ifile blocks on next write */ 906 u_int32_t lfs_nactive; /* Number of segments since last ckp */ 907 int8_t lfs_fmod; /* super block modified flag */ 908 int8_t lfs_ronly; /* mounted read-only flag */ 909 #define LFS_NOTYET 0x01 910 #define LFS_IFDIRTY 0x02 911 #define LFS_WARNED 0x04 912 #define LFS_UNDIROP 0x08 913 int8_t lfs_flags; /* currently unused flag */ 914 u_int16_t lfs_activesb; /* toggle between superblocks */ 915 daddr_t lfs_sbactive; /* disk address of current sb write */ 916 struct vnode *lfs_flushvp; /* vnode being flushed */ 917 int lfs_flushvp_fakevref; /* fake vref count for flushvp */ 918 struct vnode *lfs_unlockvp; /* being inactivated in lfs_segunlock */ 919 u_int32_t lfs_diropwait; /* # procs waiting on dirop flush */ 920 size_t lfs_devbsize; /* Device block size */ 921 size_t lfs_devbshift; /* Device block shift */ 922 krwlock_t lfs_fraglock; 923 krwlock_t lfs_iflock; /* Ifile lock */ 924 kcondvar_t lfs_stopcv; /* Wrap lock */ 925 struct lwp *lfs_stoplwp; 926 pid_t lfs_rfpid; /* Process ID of roll-forward agent */ 927 int lfs_nadirop; /* number of active dirop nodes */ 928 long lfs_ravail; /* blocks pre-reserved for writing */ 929 long lfs_favail; /* blocks pre-reserved for writing */ 930 res_t *lfs_resblk; /* Reserved memory for pageout */ 931 TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */ 932 TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */ 933 #define LFS_RESHASH_WIDTH 17 934 LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH]; 935 int lfs_pdflush; /* pagedaemon wants us to flush */ 936 u_int32_t **lfs_suflags; /* Segment use flags */ 937 #ifdef _KERNEL 938 struct pool lfs_clpool; /* Pool for struct lfs_cluster */ 939 struct pool lfs_bpppool; /* Pool for bpp */ 940 struct pool lfs_segpool; /* Pool for struct segment */ 941 #endif /* _KERNEL */ 942 #define LFS_MAX_CLEANIND 64 943 int32_t lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */ 944 int lfs_cleanind; /* Index into intervals */ 945 int lfs_sleepers; /* # procs sleeping this fs */ 946 int lfs_pages; /* dirty pages blaming this fs */ 947 lfs_bm_t *lfs_ino_bitmap; /* Inuse inodes bitmap */ 948 int lfs_nowrap; /* Suspend log wrap */ 949 int lfs_wrappass; /* Allow first log wrap requester to pass */ 950 int lfs_wrapstatus; /* Wrap status */ 951 int lfs_reclino; /* Inode being reclaimed */ 952 int lfs_startseg; /* Segment we started writing at */ 953 LIST_HEAD(, segdelta) lfs_segdhd; /* List of pending trunc accounting events */ 954 }; 955 956 /* NINDIR is the number of indirects in a file system block. */ 957 #define NINDIR(fs) ((fs)->lfs_nindir) 958 959 /* INOPB is the number of inodes in a secondary storage block. */ 960 #define INOPB(fs) ((fs)->lfs_inopb) 961 /* INOPF is the number of inodes in a fragment. */ 962 #define INOPF(fs) ((fs)->lfs_inopf) 963 964 #define blksize(fs, ip, lbn) \ 965 (((lbn) >= ULFS_NDADDR || (ip)->i_ffs1_size >= ((lbn) + 1) << (fs)->lfs_bshift) \ 966 ? (fs)->lfs_bsize \ 967 : (fragroundup(fs, blkoff(fs, (ip)->i_ffs1_size)))) 968 #define blkoff(fs, loc) ((int)((loc) & (fs)->lfs_bmask)) 969 #define fragoff(fs, loc) /* calculates (loc % fs->lfs_fsize) */ \ 970 ((int)((loc) & (fs)->lfs_ffmask)) 971 972 #if defined (_KERNEL) 973 #define fsbtodb(fs, b) ((b) << ((fs)->lfs_ffshift - DEV_BSHIFT)) 974 #define dbtofsb(fs, b) ((b) >> ((fs)->lfs_ffshift - DEV_BSHIFT)) 975 #else 976 #define fsbtodb(fs, b) ((b) << (fs)->lfs_fsbtodb) 977 #define dbtofsb(fs, b) ((b) >> (fs)->lfs_fsbtodb) 978 #endif 979 980 #define lblkno(fs, loc) ((loc) >> (fs)->lfs_bshift) 981 #define lblktosize(fs, blk) ((blk) << (fs)->lfs_bshift) 982 983 #define fsbtob(fs, b) ((b) << (fs)->lfs_ffshift) 984 #define btofsb(fs, b) ((b) >> (fs)->lfs_ffshift) 985 986 #define numfrags(fs, loc) /* calculates (loc / fs->lfs_fsize) */ \ 987 ((loc) >> (fs)->lfs_ffshift) 988 #define blkroundup(fs, size) /* calculates roundup(size, fs->lfs_bsize) */ \ 989 ((off_t)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask))) 990 #define fragroundup(fs, size) /* calculates roundup(size, fs->lfs_fsize) */ \ 991 ((off_t)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask))) 992 #define fragstoblks(fs, frags)/* calculates (frags / fs->fs_frag) */ \ 993 ((frags) >> (fs)->lfs_fbshift) 994 #define blkstofrags(fs, blks) /* calculates (blks * fs->fs_frag) */ \ 995 ((blks) << (fs)->lfs_fbshift) 996 #define fragnum(fs, fsb) /* calculates (fsb % fs->lfs_frag) */ \ 997 ((fsb) & ((fs)->lfs_frag - 1)) 998 #define blknum(fs, fsb) /* calculates rounddown(fsb, fs->lfs_frag) */ \ 999 ((fsb) &~ ((fs)->lfs_frag - 1)) 1000 #define dblksize(fs, dp, lbn) \ 1001 (((lbn) >= ULFS_NDADDR || (dp)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\ 1002 ? (fs)->lfs_bsize \ 1003 : (fragroundup(fs, blkoff(fs, (dp)->di_size)))) 1004 1005 #define segsize(fs) ((fs)->lfs_version == 1 ? \ 1006 lblktosize((fs), (fs)->lfs_ssize) : \ 1007 (fs)->lfs_ssize) 1008 #define segtod(fs, seg) (((fs)->lfs_version == 1 ? \ 1009 (fs)->lfs_ssize << (fs)->lfs_blktodb : \ 1010 btofsb((fs), (fs)->lfs_ssize)) * (seg)) 1011 #define dtosn(fs, daddr) /* block address to segment number */ \ 1012 ((uint32_t)(((daddr) - (fs)->lfs_start) / segtod((fs), 1))) 1013 #define sntod(fs, sn) /* segment number to disk address */ \ 1014 ((daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start)) 1015 1016 /* 1017 * Structures used by lfs_bmapv and lfs_markv to communicate information 1018 * about inodes and data blocks. 1019 */ 1020 typedef struct block_info { 1021 u_int32_t bi_inode; /* inode # */ 1022 int32_t bi_lbn; /* logical block w/in file */ 1023 int32_t bi_daddr; /* disk address of block */ 1024 u_int64_t bi_segcreate; /* origin segment create time */ 1025 int bi_version; /* file version number */ 1026 void *bi_bp; /* data buffer */ 1027 int bi_size; /* size of the block (if fragment) */ 1028 } BLOCK_INFO; 1029 1030 /* Compatibility for 1.5 binaries */ 1031 typedef struct block_info_15 { 1032 u_int32_t bi_inode; /* inode # */ 1033 int32_t bi_lbn; /* logical block w/in file */ 1034 int32_t bi_daddr; /* disk address of block */ 1035 u_int32_t bi_segcreate; /* origin segment create time */ 1036 int bi_version; /* file version number */ 1037 void *bi_bp; /* data buffer */ 1038 int bi_size; /* size of the block (if fragment) */ 1039 } BLOCK_INFO_15; 1040 1041 /* In-memory description of a segment about to be written. */ 1042 struct segment { 1043 struct lfs *fs; /* file system pointer */ 1044 struct buf **bpp; /* pointer to buffer array */ 1045 struct buf **cbpp; /* pointer to next available bp */ 1046 struct buf **start_bpp; /* pointer to first bp in this set */ 1047 struct buf *ibp; /* buffer pointer to inode page */ 1048 struct ulfs1_dinode *idp; /* pointer to ifile dinode */ 1049 struct finfo *fip; /* current fileinfo pointer */ 1050 struct vnode *vp; /* vnode being gathered */ 1051 void *segsum; /* segment summary info */ 1052 u_int32_t ninodes; /* number of inodes in this segment */ 1053 int32_t seg_bytes_left; /* bytes left in segment */ 1054 int32_t sum_bytes_left; /* bytes left in summary block */ 1055 u_int32_t seg_number; /* number of this segment */ 1056 int32_t *start_lbp; /* beginning lbn for this set */ 1057 1058 #define SEGM_CKP 0x0001 /* doing a checkpoint */ 1059 #define SEGM_CLEAN 0x0002 /* cleaner call; don't sort */ 1060 #define SEGM_SYNC 0x0004 /* wait for segment */ 1061 #define SEGM_PROT 0x0008 /* don't inactivate at segunlock */ 1062 #define SEGM_PAGEDAEMON 0x0010 /* pagedaemon called us */ 1063 #define SEGM_WRITERD 0x0020 /* LFS writed called us */ 1064 #define SEGM_FORCE_CKP 0x0040 /* Force checkpoint right away */ 1065 #define SEGM_RECLAIM 0x0080 /* Writing to reclaim vnode */ 1066 #define SEGM_SINGLE 0x0100 /* Opportunistic writevnodes */ 1067 u_int16_t seg_flags; /* run-time flags for this segment */ 1068 u_int32_t seg_iocount; /* number of ios pending */ 1069 int ndupino; /* number of duplicate inodes */ 1070 }; 1071 1072 #ifdef _KERNEL 1073 struct lfs_cluster { 1074 size_t bufsize; /* Size of kept data */ 1075 struct buf **bpp; /* Array of kept buffers */ 1076 int bufcount; /* Number of kept buffers */ 1077 #define LFS_CL_MALLOC 0x00000001 1078 #define LFS_CL_SHIFT 0x00000002 1079 #define LFS_CL_SYNC 0x00000004 1080 u_int32_t flags; /* Flags */ 1081 struct lfs *fs; /* LFS that this belongs to */ 1082 struct segment *seg; /* Segment structure, for LFS_CL_SYNC */ 1083 }; 1084 1085 /* 1086 * Splay tree containing block numbers allocated through lfs_balloc. 1087 */ 1088 struct lbnentry { 1089 SPLAY_ENTRY(lbnentry) entry; 1090 daddr_t lbn; 1091 }; 1092 #endif /* _KERNEL */ 1093 1094 /* 1095 * LFS inode extensions. 1096 */ 1097 struct lfs_inode_ext { 1098 off_t lfs_osize; /* size of file on disk */ 1099 u_int32_t lfs_effnblocks; /* number of blocks when i/o completes */ 1100 size_t lfs_fragsize[ULFS_NDADDR]; /* size of on-disk direct blocks */ 1101 TAILQ_ENTRY(inode) lfs_dchain; /* Dirop chain. */ 1102 TAILQ_ENTRY(inode) lfs_pchain; /* Paging chain. */ 1103 #define LFSI_NO_GOP_WRITE 0x01 1104 #define LFSI_DELETED 0x02 1105 #define LFSI_WRAPBLOCK 0x04 1106 #define LFSI_WRAPWAIT 0x08 1107 #define LFSI_BMAP 0x10 1108 u_int32_t lfs_iflags; /* Inode flags */ 1109 daddr_t lfs_hiblk; /* Highest lbn held by inode */ 1110 #ifdef _KERNEL 1111 SPLAY_HEAD(lfs_splay, lbnentry) lfs_lbtree; /* Tree of balloc'd lbns */ 1112 int lfs_nbtree; /* Size of tree */ 1113 LIST_HEAD(, segdelta) lfs_segdhd; 1114 #endif 1115 int16_t lfs_odnlink; /* on-disk nlink count for cleaner */ 1116 }; 1117 #define i_lfs_osize inode_ext.lfs->lfs_osize 1118 #define i_lfs_effnblks inode_ext.lfs->lfs_effnblocks 1119 #define i_lfs_fragsize inode_ext.lfs->lfs_fragsize 1120 #define i_lfs_dchain inode_ext.lfs->lfs_dchain 1121 #define i_lfs_pchain inode_ext.lfs->lfs_pchain 1122 #define i_lfs_iflags inode_ext.lfs->lfs_iflags 1123 #define i_lfs_hiblk inode_ext.lfs->lfs_hiblk 1124 #define i_lfs_lbtree inode_ext.lfs->lfs_lbtree 1125 #define i_lfs_nbtree inode_ext.lfs->lfs_nbtree 1126 #define i_lfs_segdhd inode_ext.lfs->lfs_segdhd 1127 #define i_lfs_odnlink inode_ext.lfs->lfs_odnlink 1128 1129 /* 1130 * Macros for determining free space on the disk, with the variable metadata 1131 * of segment summaries and inode blocks taken into account. 1132 */ 1133 /* 1134 * Estimate number of clean blocks not available for writing because 1135 * they will contain metadata or overhead. This is calculated as 1136 * 1137 * E = ((C * M / D) * D + (0) * (T - D)) / T 1138 * or more simply 1139 * E = (C * M) / T 1140 * 1141 * where 1142 * C is the clean space, 1143 * D is the dirty space, 1144 * M is the dirty metadata, and 1145 * T = C + D is the total space on disk. 1146 * 1147 * This approximates the old formula of E = C * M / D when D is close to T, 1148 * but avoids falsely reporting "disk full" when the sample size (D) is small. 1149 */ 1150 #define LFS_EST_CMETA(F) (int32_t)(( \ 1151 ((F)->lfs_dmeta * (int64_t)(F)->lfs_nclean) / \ 1152 ((F)->lfs_nseg))) 1153 1154 /* Estimate total size of the disk not including metadata */ 1155 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F)) 1156 1157 /* Estimate number of blocks actually available for writing */ 1158 #define LFS_EST_BFREE(F) ((F)->lfs_bfree > LFS_EST_CMETA(F) ? \ 1159 (F)->lfs_bfree - LFS_EST_CMETA(F) : 0) 1160 1161 /* Amount of non-meta space not available to mortal man */ 1162 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) * \ 1163 (u_int64_t)(F)->lfs_minfree) / \ 1164 100) 1165 1166 /* Can credential C write BB blocks */ 1167 #define ISSPACE(F, BB, C) \ 1168 ((((C) == NOCRED || kauth_cred_geteuid(C) == 0) && \ 1169 LFS_EST_BFREE(F) >= (BB)) || \ 1170 (kauth_cred_geteuid(C) != 0 && IS_FREESPACE(F, BB))) 1171 1172 /* Can an ordinary user write BB blocks */ 1173 #define IS_FREESPACE(F, BB) \ 1174 (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F)) 1175 1176 /* 1177 * The minimum number of blocks to create a new inode. This is: 1178 * directory direct block (1) + ULFS_NIADDR indirect blocks + inode block (1) + 1179 * ifile direct block (1) + ULFS_NIADDR indirect blocks = 3 + 2 * ULFS_NIADDR blocks. 1180 */ 1181 #define LFS_NRESERVE(F) (btofsb((F), (2 * ULFS_NIADDR + 3) << (F)->lfs_bshift)) 1182 1183 /* Statistics Counters */ 1184 struct lfs_stats { /* Must match sysctl list in lfs_vfsops.h ! */ 1185 u_int segsused; 1186 u_int psegwrites; 1187 u_int psyncwrites; 1188 u_int pcleanwrites; 1189 u_int blocktot; 1190 u_int cleanblocks; 1191 u_int ncheckpoints; 1192 u_int nwrites; 1193 u_int nsync_writes; 1194 u_int wait_exceeded; 1195 u_int write_exceeded; 1196 u_int flush_invoked; 1197 u_int vflush_invoked; 1198 u_int clean_inlocked; 1199 u_int clean_vnlocked; 1200 u_int segs_reclaimed; 1201 }; 1202 #ifdef _KERNEL 1203 extern struct lfs_stats lfs_stats; 1204 #endif 1205 1206 /* Fcntls to take the place of the lfs syscalls */ 1207 struct lfs_fcntl_markv { 1208 BLOCK_INFO *blkiov; /* blocks to relocate */ 1209 int blkcnt; /* number of blocks */ 1210 }; 1211 1212 #define LFCNSEGWAITALL _FCNR_FSPRIV('L', 14, struct timeval) 1213 #define LFCNSEGWAIT _FCNR_FSPRIV('L', 15, struct timeval) 1214 #define LFCNBMAPV _FCNRW_FSPRIV('L', 2, struct lfs_fcntl_markv) 1215 #define LFCNMARKV _FCNRW_FSPRIV('L', 3, struct lfs_fcntl_markv) 1216 #define LFCNRECLAIM _FCNO_FSPRIV('L', 4) 1217 1218 struct lfs_fhandle { 1219 char space[28]; /* FHANDLE_SIZE_COMPAT (but used from userland too) */ 1220 }; 1221 #define LFCNREWIND _FCNR_FSPRIV('L', 6, int) 1222 #define LFCNINVAL _FCNR_FSPRIV('L', 7, int) 1223 #define LFCNRESIZE _FCNR_FSPRIV('L', 8, int) 1224 #define LFCNWRAPSTOP _FCNR_FSPRIV('L', 9, int) 1225 #define LFCNWRAPGO _FCNR_FSPRIV('L', 10, int) 1226 #define LFCNIFILEFH _FCNW_FSPRIV('L', 11, struct lfs_fhandle) 1227 #define LFCNWRAPPASS _FCNR_FSPRIV('L', 12, int) 1228 # define LFS_WRAP_GOING 0x0 1229 # define LFS_WRAP_WAITING 0x1 1230 #define LFCNWRAPSTATUS _FCNW_FSPRIV('L', 13, int) 1231 1232 /* 1233 * Compat. Defined for kernel only. Userland always uses 1234 * "the one true version". 1235 */ 1236 #ifdef _KERNEL 1237 #include <compat/sys/time_types.h> 1238 1239 #define LFCNSEGWAITALL_COMPAT _FCNW_FSPRIV('L', 0, struct timeval50) 1240 #define LFCNSEGWAIT_COMPAT _FCNW_FSPRIV('L', 1, struct timeval50) 1241 #define LFCNIFILEFH_COMPAT _FCNW_FSPRIV('L', 5, struct lfs_fhandle) 1242 #define LFCNIFILEFH_COMPAT2 _FCN_FSPRIV(F_FSOUT, 'L', 11, 32) 1243 #define LFCNWRAPSTOP_COMPAT _FCNO_FSPRIV('L', 9) 1244 #define LFCNWRAPGO_COMPAT _FCNO_FSPRIV('L', 10) 1245 #define LFCNSEGWAITALL_COMPAT_50 _FCNR_FSPRIV('L', 0, struct timeval50) 1246 #define LFCNSEGWAIT_COMPAT_50 _FCNR_FSPRIV('L', 1, struct timeval50) 1247 #endif 1248 1249 #ifdef _KERNEL 1250 /* XXX MP */ 1251 #define LFS_SEGLOCK_HELD(fs) \ 1252 ((fs)->lfs_seglock != 0 && \ 1253 (fs)->lfs_lockpid == curproc->p_pid && \ 1254 (fs)->lfs_locklwp == curlwp->l_lid) 1255 #endif /* _KERNEL */ 1256 1257 /* Debug segment lock */ 1258 #ifdef notyet 1259 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs)) 1260 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs)) 1261 # define ASSERT_DUNNO_SEGLOCK(fs) 1262 # define ASSERT_MAYBE_SEGLOCK(fs) 1263 #else /* !notyet */ 1264 # define ASSERT_DUNNO_SEGLOCK(fs) \ 1265 DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \ 1266 LFS_SEGLOCK_HELD(fs))) 1267 # define ASSERT_SEGLOCK(fs) do { \ 1268 if (!LFS_SEGLOCK_HELD(fs)) { \ 1269 DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \ 1270 } \ 1271 } while(0) 1272 # define ASSERT_NO_SEGLOCK(fs) do { \ 1273 if (LFS_SEGLOCK_HELD(fs)) { \ 1274 DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \ 1275 } \ 1276 } while(0) 1277 # define ASSERT_MAYBE_SEGLOCK(x) 1278 #endif /* !notyet */ 1279 1280 __BEGIN_DECLS 1281 void lfs_itimes(struct inode *, const struct timespec *, 1282 const struct timespec *, const struct timespec *); 1283 __END_DECLS 1284 1285 #endif /* !_UFS_LFS_LFS_H_ */ 1286
Indexes created Tue Sep 30 11:09:46 GMT 2025