lfs.h revision 1.129
1 /* $NetBSD: lfs.h,v 1.129 2009/10/29 17:10:32 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Konrad E. Schroder <perseant (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 #ifdef COMPAT_50 70 #include <compat/sys/time.h> 71 #endif 72 73 /* 74 * Compile-time options for LFS. 75 */ 76 #define LFS_IFIND_RETRIES 16 77 #define LFS_LOGLENGTH 1024 /* size of debugging log */ 78 #define LFS_MAX_ACTIVE 10 /* Dirty segments before ckp forced */ 79 80 /* 81 * Fixed filesystem layout parameters 82 */ 83 #define LFS_LABELPAD 8192 /* LFS label size */ 84 #define LFS_SBPAD 8192 /* LFS superblock size */ 85 86 #define LFS_UNUSED_INUM 0 /* 0: out of band inode number */ 87 #define LFS_IFILE_INUM 1 /* 1: IFILE inode number */ 88 /* 2: Root inode number */ 89 #define LOSTFOUNDINO 3 /* 3: lost+found inode number */ 90 #define LFS_FIRST_INUM 4 /* 4: first free inode number */ 91 92 #define LFS_V1_SUMMARY_SIZE 512 /* V1 fixed summary size */ 93 #define LFS_DFL_SUMMARY_SIZE 512 /* Default summary size */ 94 95 #define LFS_MAX_DADDR 0x7fffffff /* Highest addressable fsb */ 96 97 #define LFS_MAXNAMLEN 255 /* maximum name length in a dir */ 98 99 /* Adjustable filesystem parameters */ 100 #define MIN_FREE_SEGS 20 101 #define MIN_RESV_SEGS 15 102 #ifndef LFS_ATIME_IFILE 103 # define LFS_ATIME_IFILE 0 /* Store atime info in ifile (optional in LFSv1) */ 104 #endif 105 #define LFS_MARKV_MAXBLKCNT 65536 /* Max block count for lfs_markv() */ 106 107 /* Misc. definitions */ 108 #define BW_CLEAN 1 /* Flag for lfs_bwrite_ext() */ 109 #define PG_DELWRI PG_PAGER1 /* Local def for delayed pageout */ 110 111 /* Resource limits */ 112 #define LFS_MAX_RESOURCE(x, u) (((x) >> 2) - 10 * (u)) 113 #define LFS_WAIT_RESOURCE(x, u) (((x) >> 1) - ((x) >> 3) - 10 * (u)) 114 #define LFS_INVERSE_MAX_RESOURCE(x, u) (((x) + 10 * (u)) << 2) 115 #define LFS_MAX_BUFS LFS_MAX_RESOURCE(nbuf, 1) 116 #define LFS_WAIT_BUFS LFS_WAIT_RESOURCE(nbuf, 1) 117 #define LFS_INVERSE_MAX_BUFS(n) LFS_INVERSE_MAX_RESOURCE(n, 1) 118 #define LFS_MAX_BYTES LFS_MAX_RESOURCE(bufmem_lowater, PAGE_SIZE) 119 #define LFS_INVERSE_MAX_BYTES(n) LFS_INVERSE_MAX_RESOURCE(n, PAGE_SIZE) 120 #define LFS_WAIT_BYTES LFS_WAIT_RESOURCE(bufmem_lowater, PAGE_SIZE) 121 #define LFS_MAX_DIROP ((desiredvnodes >> 2) + (desiredvnodes >> 3)) 122 #define SIZEOF_DIROP(fs) (2 * ((fs)->lfs_bsize + DINODE1_SIZE)) 123 #define LFS_MAX_FSDIROP(fs) \ 124 ((fs)->lfs_nclean <= (fs)->lfs_resvseg ? 0 : \ 125 (((fs)->lfs_nclean - (fs)->lfs_resvseg) * (fs)->lfs_ssize) / \ 126 (2 * SIZEOF_DIROP(fs))) 127 #define LFS_MAX_PAGES lfs_max_pages() 128 #define LFS_WAIT_PAGES lfs_wait_pages() 129 #define LFS_BUFWAIT 2 /* How long to wait if over *_WAIT_* */ 130 131 #ifdef _KERNEL 132 int lfs_wait_pages(void); 133 int lfs_max_pages(void); 134 #endif /* _KERNEL */ 135 136 /* How starved can we be before we start holding back page writes */ 137 #define LFS_STARVED_FOR_SEGS(fs) ((fs)->lfs_nclean < (fs)->lfs_resvseg) 138 139 /* 140 * Reserved blocks for lfs_malloc 141 */ 142 143 /* Structure to keep reserved blocks */ 144 typedef struct lfs_res_blk { 145 void *p; 146 LIST_ENTRY(lfs_res_blk) res; 147 int size; 148 char inuse; 149 } res_t; 150 151 /* Types for lfs_newbuf and lfs_malloc */ 152 #define LFS_NB_UNKNOWN -1 153 #define LFS_NB_SUMMARY 0 154 #define LFS_NB_SBLOCK 1 155 #define LFS_NB_IBLOCK 2 156 #define LFS_NB_CLUSTER 3 157 #define LFS_NB_CLEAN 4 158 #define LFS_NB_BLKIOV 5 159 #define LFS_NB_COUNT 6 /* always last */ 160 161 /* Number of reserved memory blocks of each type */ 162 #define LFS_N_SUMMARIES 2 163 #define LFS_N_SBLOCKS 1 /* Always 1, to throttle superblock writes */ 164 #define LFS_N_IBLOCKS 16 /* In theory ssize/bsize; in practice around 2 */ 165 #define LFS_N_CLUSTERS 16 /* In theory ssize/MAXPHYS */ 166 #define LFS_N_CLEAN 0 167 #define LFS_N_BLKIOV 1 168 169 /* Total count of "large" (non-pool) types */ 170 #define LFS_N_TOTAL (LFS_N_SUMMARIES + LFS_N_SBLOCKS + LFS_N_IBLOCKS + \ 171 LFS_N_CLUSTERS + LFS_N_CLEAN + LFS_N_BLKIOV) 172 173 /* Counts for pool types */ 174 #define LFS_N_CL LFS_N_CLUSTERS 175 #define LFS_N_BPP 2 176 #define LFS_N_SEG 2 177 178 /* 179 * "struct buf" associated definitions 180 */ 181 182 /* Unassigned disk addresses. */ 183 #define UNASSIGNED -1 184 #define UNWRITTEN -2 185 186 /* Unused logical block number */ 187 #define LFS_UNUSED_LBN -1 188 189 /* Determine if a buffer belongs to the ifile */ 190 #define IS_IFILE(bp) (VTOI(bp->b_vp)->i_number == LFS_IFILE_INUM) 191 192 # define LFS_LOCK_BUF(bp) do { \ 193 if (((bp)->b_flags & B_LOCKED) == 0 && bp->b_iodone == NULL) { \ 194 mutex_enter(&lfs_lock); \ 195 ++locked_queue_count; \ 196 locked_queue_bytes += bp->b_bufsize; \ 197 mutex_exit(&lfs_lock); \ 198 } \ 199 (bp)->b_flags |= B_LOCKED; \ 200 } while (0) 201 202 # define LFS_UNLOCK_BUF(bp) do { \ 203 if (((bp)->b_flags & B_LOCKED) != 0 && bp->b_iodone == NULL) { \ 204 mutex_enter(&lfs_lock); \ 205 --locked_queue_count; \ 206 locked_queue_bytes -= bp->b_bufsize; \ 207 if (locked_queue_count < LFS_WAIT_BUFS && \ 208 locked_queue_bytes < LFS_WAIT_BYTES) \ 209 wakeup(&locked_queue_count); \ 210 mutex_exit(&lfs_lock); \ 211 } \ 212 (bp)->b_flags &= ~B_LOCKED; \ 213 } while (0) 214 215 #ifdef _KERNEL 216 217 extern u_long bufmem_lowater, bufmem_hiwater; /* XXX */ 218 219 # define LFS_IS_MALLOC_BUF(bp) ((bp)->b_iodone == lfs_callback) 220 221 # ifdef DEBUG 222 # define LFS_DEBUG_COUNTLOCKED(m) do { \ 223 if (lfs_debug_log_subsys[DLOG_LLIST]) { \ 224 lfs_countlocked(&locked_queue_count, &locked_queue_bytes, (m)); \ 225 wakeup(&locked_queue_count); \ 226 } \ 227 } while (0) 228 # else 229 # define LFS_DEBUG_COUNTLOCKED(m) 230 # endif 231 232 /* log for debugging writes to the Ifile */ 233 # ifdef DEBUG 234 struct lfs_log_entry { 235 const char *op; 236 const char *file; 237 int pid; 238 int line; 239 daddr_t block; 240 unsigned long flags; 241 }; 242 extern int lfs_lognum; 243 extern struct lfs_log_entry lfs_log[LFS_LOGLENGTH]; 244 # define LFS_BWRITE_LOG(bp) lfs_bwrite_log((bp), __FILE__, __LINE__) 245 # define LFS_ENTER_LOG(theop, thefile, theline, lbn, theflags, thepid) do {\ 246 int _s; \ 247 \ 248 mutex_enter(&lfs_lock); \ 249 _s = splbio(); \ 250 lfs_log[lfs_lognum].op = theop; \ 251 lfs_log[lfs_lognum].file = thefile; \ 252 lfs_log[lfs_lognum].line = (theline); \ 253 lfs_log[lfs_lognum].pid = (thepid); \ 254 lfs_log[lfs_lognum].block = (lbn); \ 255 lfs_log[lfs_lognum].flags = (theflags); \ 256 lfs_lognum = (lfs_lognum + 1) % LFS_LOGLENGTH; \ 257 splx(_s); \ 258 mutex_exit(&lfs_lock); \ 259 } while (0) 260 261 # define LFS_BCLEAN_LOG(fs, bp) do { \ 262 if ((bp)->b_vp == (fs)->lfs_ivnode) \ 263 LFS_ENTER_LOG("clear", __FILE__, __LINE__, \ 264 bp->b_lblkno, bp->b_flags, curproc->p_pid);\ 265 } while (0) 266 267 /* Must match list in lfs_vfsops.c ! */ 268 # define DLOG_RF 0 /* roll forward */ 269 # define DLOG_ALLOC 1 /* inode alloc */ 270 # define DLOG_AVAIL 2 /* lfs_{,r,f}avail */ 271 # define DLOG_FLUSH 3 /* flush */ 272 # define DLOG_LLIST 4 /* locked list accounting */ 273 # define DLOG_WVNODE 5 /* vflush/writevnodes verbose */ 274 # define DLOG_VNODE 6 /* vflush/writevnodes */ 275 # define DLOG_SEG 7 /* segwrite */ 276 # define DLOG_SU 8 /* seguse accounting */ 277 # define DLOG_CLEAN 9 /* cleaner routines */ 278 # define DLOG_MOUNT 10 /* mount/unmount */ 279 # define DLOG_PAGE 11 /* putpages/gop_write */ 280 # define DLOG_DIROP 12 /* dirop accounting */ 281 # define DLOG_MALLOC 13 /* lfs_malloc accounting */ 282 # define DLOG_MAX 14 /* The terminator */ 283 # define DLOG(a) lfs_debug_log a 284 # else /* ! DEBUG */ 285 # define LFS_BCLEAN_LOG(fs, bp) 286 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)) 287 # define DLOG(a) 288 # endif /* ! DEBUG */ 289 #else /* ! _KERNEL */ 290 # define LFS_BWRITE_LOG(bp) VOP_BWRITE((bp)) 291 #endif /* _KERNEL */ 292 293 #ifdef _KERNEL 294 /* Filehandle structure for exported LFSes */ 295 struct lfid { 296 struct ufid lfid_ufid; 297 #define lfid_len lfid_ufid.ufid_len 298 #define lfid_ino lfid_ufid.ufid_ino 299 #define lfid_gen lfid_ufid.ufid_gen 300 uint32_t lfid_ident; 301 }; 302 #endif /* _KERNEL */ 303 304 /* 305 * "struct inode" associated definitions 306 */ 307 308 /* Address calculations for metadata located in the inode */ 309 #define S_INDIR(fs) -NDADDR 310 #define D_INDIR(fs) (S_INDIR(fs) - NINDIR(fs) - 1) 311 #define T_INDIR(fs) (D_INDIR(fs) - NINDIR(fs) * NINDIR(fs) - 1) 312 313 /* For convenience */ 314 #define IN_ALLMOD (IN_MODIFIED|IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY|IN_ACCESSED|IN_CLEANING) 315 316 #define LFS_SET_UINO(ip, flags) do { \ 317 if (((flags) & IN_ACCESSED) && !((ip)->i_flag & IN_ACCESSED)) \ 318 ++(ip)->i_lfs->lfs_uinodes; \ 319 if (((flags) & IN_CLEANING) && !((ip)->i_flag & IN_CLEANING)) \ 320 ++(ip)->i_lfs->lfs_uinodes; \ 321 if (((flags) & IN_MODIFIED) && !((ip)->i_flag & IN_MODIFIED)) \ 322 ++(ip)->i_lfs->lfs_uinodes; \ 323 (ip)->i_flag |= (flags); \ 324 } while (0) 325 326 #define LFS_CLR_UINO(ip, flags) do { \ 327 if (((flags) & IN_ACCESSED) && ((ip)->i_flag & IN_ACCESSED)) \ 328 --(ip)->i_lfs->lfs_uinodes; \ 329 if (((flags) & IN_CLEANING) && ((ip)->i_flag & IN_CLEANING)) \ 330 --(ip)->i_lfs->lfs_uinodes; \ 331 if (((flags) & IN_MODIFIED) && ((ip)->i_flag & IN_MODIFIED)) \ 332 --(ip)->i_lfs->lfs_uinodes; \ 333 (ip)->i_flag &= ~(flags); \ 334 if ((ip)->i_lfs->lfs_uinodes < 0) { \ 335 panic("lfs_uinodes < 0"); \ 336 } \ 337 } while (0) 338 339 #define LFS_ITIMES(ip, acc, mod, cre) \ 340 while ((ip)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY)) \ 341 lfs_itimes(ip, acc, mod, cre) 342 343 /* 344 * "struct vnode" associated definitions 345 */ 346 347 /* Heuristic emptiness measure */ 348 #define VPISEMPTY(vp) (LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 349 !(vp->v_type == VREG && (vp)->v_iflag & VI_ONWORKLST) &&\ 350 VTOI(vp)->i_lfs_nbtree == 0) 351 352 #define WRITEINPROG(vp) ((vp)->v_numoutput > 0 || \ 353 (!LIST_EMPTY(&(vp)->v_dirtyblkhd) && \ 354 !(VTOI(vp)->i_flag & (IN_MODIFIED | IN_ACCESSED | IN_CLEANING)))) 355 356 357 /* 358 * On-disk and in-memory checkpoint segment usage structure. 359 */ 360 typedef struct segusage SEGUSE; 361 struct segusage { 362 u_int32_t su_nbytes; /* 0: number of live bytes */ 363 u_int32_t su_olastmod; /* 4: SEGUSE last modified timestamp */ 364 u_int16_t su_nsums; /* 8: number of summaries in segment */ 365 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 366 367 #define SEGUSE_ACTIVE 0x01 /* segment currently being written */ 368 #define SEGUSE_DIRTY 0x02 /* segment has data in it */ 369 #define SEGUSE_SUPERBLOCK 0x04 /* segment contains a superblock */ 370 #define SEGUSE_ERROR 0x08 /* cleaner: do not clean segment */ 371 #define SEGUSE_EMPTY 0x10 /* segment is empty */ 372 #define SEGUSE_INVAL 0x20 /* segment is invalid */ 373 u_int32_t su_flags; /* 12: segment flags */ 374 u_int64_t su_lastmod; /* 16: last modified timestamp */ 375 }; 376 377 typedef struct segusage_v1 SEGUSE_V1; 378 struct segusage_v1 { 379 u_int32_t su_nbytes; /* 0: number of live bytes */ 380 u_int32_t su_lastmod; /* 4: SEGUSE last modified timestamp */ 381 u_int16_t su_nsums; /* 8: number of summaries in segment */ 382 u_int16_t su_ninos; /* 10: number of inode blocks in seg */ 383 u_int32_t su_flags; /* 12: segment flags */ 384 }; 385 386 #define SEGUPB(fs) (fs->lfs_sepb) 387 #define SEGTABSIZE_SU(fs) \ 388 (((fs)->lfs_nseg + SEGUPB(fs) - 1) / (fs)->lfs_sepb) 389 390 #ifdef _KERNEL 391 # define SHARE_IFLOCK(F) \ 392 do { \ 393 rw_enter(&(F)->lfs_iflock, RW_READER); \ 394 } while(0) 395 # define UNSHARE_IFLOCK(F) \ 396 do { \ 397 rw_exit(&(F)->lfs_iflock); \ 398 } while(0) 399 #else /* ! _KERNEL */ 400 # define SHARE_IFLOCK(F) 401 # define UNSHARE_IFLOCK(F) 402 #endif /* ! _KERNEL */ 403 404 /* Read in the block with a specific segment usage entry from the ifile. */ 405 #define LFS_SEGENTRY(SP, F, IN, BP) do { \ 406 int _e; \ 407 SHARE_IFLOCK(F); \ 408 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 409 if ((_e = bread((F)->lfs_ivnode, \ 410 ((IN) / (F)->lfs_sepb) + (F)->lfs_cleansz, \ 411 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 412 panic("lfs: ifile read: %d", _e); \ 413 if ((F)->lfs_version == 1) \ 414 (SP) = (SEGUSE *)((SEGUSE_V1 *)(BP)->b_data + \ 415 ((IN) & ((F)->lfs_sepb - 1))); \ 416 else \ 417 (SP) = (SEGUSE *)(BP)->b_data + ((IN) % (F)->lfs_sepb); \ 418 UNSHARE_IFLOCK(F); \ 419 } while (0) 420 421 #define LFS_WRITESEGENTRY(SP, F, IN, BP) do { \ 422 if ((SP)->su_nbytes == 0) \ 423 (SP)->su_flags |= SEGUSE_EMPTY; \ 424 else \ 425 (SP)->su_flags &= ~SEGUSE_EMPTY; \ 426 (F)->lfs_suflags[(F)->lfs_activesb][(IN)] = (SP)->su_flags; \ 427 LFS_BWRITE_LOG(BP); \ 428 } while (0) 429 430 /* 431 * On-disk file information. One per file with data blocks in the segment. 432 */ 433 typedef struct finfo FINFO; 434 struct finfo { 435 u_int32_t fi_nblocks; /* number of blocks */ 436 u_int32_t fi_version; /* version number */ 437 u_int32_t fi_ino; /* inode number */ 438 u_int32_t fi_lastlength; /* length of last block in array */ 439 int32_t fi_blocks[1]; /* array of logical block numbers */ 440 }; 441 /* sizeof FINFO except fi_blocks */ 442 #define FINFOSIZE (sizeof(FINFO) - sizeof(int32_t)) 443 444 /* 445 * Index file inode entries. 446 */ 447 typedef struct ifile IFILE; 448 struct ifile { 449 u_int32_t if_version; /* inode version number */ 450 #define LFS_UNUSED_DADDR 0 /* out-of-band daddr */ 451 int32_t if_daddr; /* inode disk address */ 452 #define LFS_ORPHAN_NEXTFREE (~(u_int32_t)0) /* indicate orphaned file */ 453 u_int32_t if_nextfree; /* next-unallocated inode */ 454 u_int32_t if_atime_sec; /* Last access time, seconds */ 455 u_int32_t if_atime_nsec; /* and nanoseconds */ 456 }; 457 458 typedef struct ifile_v1 IFILE_V1; 459 struct ifile_v1 { 460 u_int32_t if_version; /* inode version number */ 461 int32_t if_daddr; /* inode disk address */ 462 u_int32_t if_nextfree; /* next-unallocated inode */ 463 #if LFS_ATIME_IFILE 464 struct timespec if_atime; /* Last access time */ 465 #endif 466 }; 467 468 /* 469 * LFSv1 compatibility code is not allowed to touch if_atime, since it 470 * may not be mapped! 471 */ 472 /* Read in the block with a specific inode from the ifile. */ 473 #define LFS_IENTRY(IP, F, IN, BP) do { \ 474 int _e; \ 475 SHARE_IFLOCK(F); \ 476 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 477 if ((_e = bread((F)->lfs_ivnode, \ 478 (IN) / (F)->lfs_ifpb + (F)->lfs_cleansz + (F)->lfs_segtabsz, \ 479 (F)->lfs_bsize, NOCRED, 0, &(BP))) != 0) \ 480 panic("lfs: ifile ino %d read %d", (int)(IN), _e); \ 481 if ((F)->lfs_version == 1) \ 482 (IP) = (IFILE *)((IFILE_V1 *)(BP)->b_data + \ 483 (IN) % (F)->lfs_ifpb); \ 484 else \ 485 (IP) = (IFILE *)(BP)->b_data + (IN) % (F)->lfs_ifpb; \ 486 UNSHARE_IFLOCK(F); \ 487 } while (0) 488 489 /* 490 * Cleaner information structure. This resides in the ifile and is used 491 * to pass information from the kernel to the cleaner. 492 */ 493 typedef struct _cleanerinfo { 494 u_int32_t clean; /* number of clean segments */ 495 u_int32_t dirty; /* number of dirty segments */ 496 int32_t bfree; /* disk blocks free */ 497 int32_t avail; /* disk blocks available */ 498 u_int32_t free_head; /* head of the inode free list */ 499 u_int32_t free_tail; /* tail of the inode free list */ 500 #define LFS_CLEANER_MUST_CLEAN 0x01 501 u_int32_t flags; /* status word from the kernel */ 502 } CLEANERINFO; 503 504 #define CLEANSIZE_SU(fs) \ 505 ((sizeof(CLEANERINFO) + (fs)->lfs_bsize - 1) >> (fs)->lfs_bshift) 506 507 /* Read in the block with the cleaner info from the ifile. */ 508 #define LFS_CLEANERINFO(CP, F, BP) do { \ 509 SHARE_IFLOCK(F); \ 510 VTOI((F)->lfs_ivnode)->i_flag |= IN_ACCESS; \ 511 if (bread((F)->lfs_ivnode, \ 512 (daddr_t)0, (F)->lfs_bsize, NOCRED, 0, &(BP))) \ 513 panic("lfs: ifile read"); \ 514 (CP) = (CLEANERINFO *)(BP)->b_data; \ 515 UNSHARE_IFLOCK(F); \ 516 } while (0) 517 518 /* 519 * Synchronize the Ifile cleaner info with current avail and bfree. 520 */ 521 #define LFS_SYNC_CLEANERINFO(cip, fs, bp, w) do { \ 522 mutex_enter(&lfs_lock); \ 523 if ((w) || (cip)->bfree != (fs)->lfs_bfree || \ 524 (cip)->avail != (fs)->lfs_avail - (fs)->lfs_ravail - \ 525 (fs)->lfs_favail) { \ 526 (cip)->bfree = (fs)->lfs_bfree; \ 527 (cip)->avail = (fs)->lfs_avail - (fs)->lfs_ravail - \ 528 (fs)->lfs_favail; \ 529 if (((bp)->b_flags & B_GATHERED) == 0) { \ 530 (fs)->lfs_flags |= LFS_IFDIRTY; \ 531 } \ 532 mutex_exit(&lfs_lock); \ 533 (void) LFS_BWRITE_LOG(bp); /* Ifile */ \ 534 } else { \ 535 mutex_exit(&lfs_lock); \ 536 brelse(bp, 0); \ 537 } \ 538 } while (0) 539 540 /* 541 * Get the head of the inode free list. 542 * Always called with the segment lock held. 543 */ 544 #define LFS_GET_HEADFREE(FS, CIP, BP, FREEP) do { \ 545 if ((FS)->lfs_version > 1) { \ 546 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 547 (FS)->lfs_freehd = (CIP)->free_head; \ 548 brelse(BP, 0); \ 549 } \ 550 *(FREEP) = (FS)->lfs_freehd; \ 551 } while (0) 552 553 #define LFS_PUT_HEADFREE(FS, CIP, BP, VAL) do { \ 554 (FS)->lfs_freehd = (VAL); \ 555 if ((FS)->lfs_version > 1) { \ 556 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 557 (CIP)->free_head = (VAL); \ 558 LFS_BWRITE_LOG(BP); \ 559 mutex_enter(&lfs_lock); \ 560 (FS)->lfs_flags |= LFS_IFDIRTY; \ 561 mutex_exit(&lfs_lock); \ 562 } \ 563 } while (0) 564 565 #define LFS_GET_TAILFREE(FS, CIP, BP, FREEP) do { \ 566 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 567 *(FREEP) = (CIP)->free_tail; \ 568 brelse(BP, 0); \ 569 } while (0) 570 571 #define LFS_PUT_TAILFREE(FS, CIP, BP, VAL) do { \ 572 LFS_CLEANERINFO((CIP), (FS), (BP)); \ 573 (CIP)->free_tail = (VAL); \ 574 LFS_BWRITE_LOG(BP); \ 575 mutex_enter(&lfs_lock); \ 576 (FS)->lfs_flags |= LFS_IFDIRTY; \ 577 mutex_exit(&lfs_lock); \ 578 } while (0) 579 580 /* 581 * On-disk segment summary information 582 */ 583 typedef struct segsum_v1 SEGSUM_V1; 584 struct segsum_v1 { 585 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 586 u_int32_t ss_datasum; /* 4: check sum of data */ 587 u_int32_t ss_magic; /* 8: segment summary magic number */ 588 #define SS_MAGIC 0x061561 589 int32_t ss_next; /* 12: next segment */ 590 u_int32_t ss_create; /* 16: creation time stamp */ 591 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 592 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 593 594 #define SS_DIROP 0x01 /* segment begins a dirop */ 595 #define SS_CONT 0x02 /* more partials to finish this write*/ 596 #define SS_CLEAN 0x04 /* written by the cleaner */ 597 #define SS_RFW 0x08 /* written by the roll-forward agent */ 598 u_int16_t ss_flags; /* 24: used for directory operations */ 599 u_int16_t ss_pad; /* 26: extra space */ 600 /* FINFO's and inode daddr's... */ 601 }; 602 603 typedef struct segsum SEGSUM; 604 struct segsum { 605 u_int32_t ss_sumsum; /* 0: check sum of summary block */ 606 u_int32_t ss_datasum; /* 4: check sum of data */ 607 u_int32_t ss_magic; /* 8: segment summary magic number */ 608 int32_t ss_next; /* 12: next segment */ 609 u_int32_t ss_ident; /* 16: roll-forward fsid */ 610 #define ss_ocreate ss_ident /* ident is where create was in v1 */ 611 u_int16_t ss_nfinfo; /* 20: number of file info structures */ 612 u_int16_t ss_ninos; /* 22: number of inodes in summary */ 613 u_int16_t ss_flags; /* 24: used for directory operations */ 614 u_int8_t ss_pad[6]; /* 26: extra space */ 615 u_int64_t ss_serial; /* 32: serial number */ 616 u_int64_t ss_create; /* 40: time stamp */ 617 /* FINFO's and inode daddr's... */ 618 }; 619 620 #define SEGSUM_SIZE(fs) ((fs)->lfs_version == 1 ? sizeof(SEGSUM_V1) : sizeof(SEGSUM)) 621 622 623 /* 624 * On-disk super block. 625 */ 626 struct dlfs { 627 #define LFS_MAGIC 0x070162 628 u_int32_t dlfs_magic; /* 0: magic number */ 629 #define LFS_VERSION 2 630 u_int32_t dlfs_version; /* 4: version number */ 631 632 u_int32_t dlfs_size; /* 8: number of blocks in fs (v1) */ 633 /* number of frags in fs (v2) */ 634 u_int32_t dlfs_ssize; /* 12: number of blocks per segment (v1) */ 635 /* number of bytes per segment (v2) */ 636 u_int32_t dlfs_dsize; /* 16: number of disk blocks in fs */ 637 u_int32_t dlfs_bsize; /* 20: file system block size */ 638 u_int32_t dlfs_fsize; /* 24: size of frag blocks in fs */ 639 u_int32_t dlfs_frag; /* 28: number of frags in a block in fs */ 640 641 /* Checkpoint region. */ 642 u_int32_t dlfs_freehd; /* 32: start of the free list */ 643 int32_t dlfs_bfree; /* 36: number of free disk blocks */ 644 u_int32_t dlfs_nfiles; /* 40: number of allocated inodes */ 645 int32_t dlfs_avail; /* 44: blocks available for writing */ 646 int32_t dlfs_uinodes; /* 48: inodes in cache not yet on disk */ 647 int32_t dlfs_idaddr; /* 52: inode file disk address */ 648 u_int32_t dlfs_ifile; /* 56: inode file inode number */ 649 int32_t dlfs_lastseg; /* 60: address of last segment written */ 650 int32_t dlfs_nextseg; /* 64: address of next segment to write */ 651 int32_t dlfs_curseg; /* 68: current segment being written */ 652 int32_t dlfs_offset; /* 72: offset in curseg for next partial */ 653 int32_t dlfs_lastpseg; /* 76: address of last partial written */ 654 u_int32_t dlfs_inopf; /* 80: v1: time stamp; v2: inodes per frag */ 655 #define dlfs_otstamp dlfs_inopf 656 657 /* These are configuration parameters. */ 658 u_int32_t dlfs_minfree; /* 84: minimum percentage of free blocks */ 659 660 /* These fields can be computed from the others. */ 661 u_int64_t dlfs_maxfilesize; /* 88: maximum representable file size */ 662 u_int32_t dlfs_fsbpseg; /* 96: fsb per segment */ 663 u_int32_t dlfs_inopb; /* 100: inodes per block */ 664 u_int32_t dlfs_ifpb; /* 104: IFILE entries per block */ 665 u_int32_t dlfs_sepb; /* 108: SEGUSE entries per block */ 666 u_int32_t dlfs_nindir; /* 112: indirect pointers per block */ 667 u_int32_t dlfs_nseg; /* 116: number of segments */ 668 u_int32_t dlfs_nspf; /* 120: number of sectors per fragment */ 669 u_int32_t dlfs_cleansz; /* 124: cleaner info size in blocks */ 670 u_int32_t dlfs_segtabsz; /* 128: segment table size in blocks */ 671 u_int32_t dlfs_segmask; /* 132: calculate offset within a segment */ 672 u_int32_t dlfs_segshift; /* 136: fast mult/div for segments */ 673 u_int32_t dlfs_bshift; /* 140: calc block number from file offset */ 674 u_int32_t dlfs_ffshift; /* 144: fast mult/div for frag from file */ 675 u_int32_t dlfs_fbshift; /* 148: fast mult/div for frag from block */ 676 u_int64_t dlfs_bmask; /* 152: calc block offset from file offset */ 677 u_int64_t dlfs_ffmask; /* 160: calc frag offset from file offset */ 678 u_int64_t dlfs_fbmask; /* 168: calc frag offset from block offset */ 679 u_int32_t dlfs_blktodb; /* 176: blktodb and dbtoblk shift constant */ 680 u_int32_t dlfs_sushift; /* 180: fast mult/div for segusage table */ 681 682 int32_t dlfs_maxsymlinklen; /* 184: max length of an internal symlink */ 683 #define LFS_MIN_SBINTERVAL 5 /* minimum superblock segment spacing */ 684 #define LFS_MAXNUMSB 10 /* 188: superblock disk offsets */ 685 int32_t dlfs_sboffs[LFS_MAXNUMSB]; 686 687 u_int32_t dlfs_nclean; /* 228: Number of clean segments */ 688 u_char dlfs_fsmnt[MNAMELEN]; /* 232: name mounted on */ 689 #define LFS_PF_CLEAN 0x1 690 u_int16_t dlfs_pflags; /* 322: file system persistent flags */ 691 int32_t dlfs_dmeta; /* 324: total number of dirty summaries */ 692 u_int32_t dlfs_minfreeseg; /* 328: segments not counted in bfree */ 693 u_int32_t dlfs_sumsize; /* 332: size of summary blocks */ 694 u_int64_t dlfs_serial; /* 336: serial number */ 695 u_int32_t dlfs_ibsize; /* 344: size of inode blocks */ 696 int32_t dlfs_start; /* 348: start of segment 0 */ 697 u_int64_t dlfs_tstamp; /* 352: time stamp */ 698 #define LFS_44INODEFMT 0 699 #define LFS_MAXINODEFMT 0 700 u_int32_t dlfs_inodefmt; /* 360: inode format version */ 701 u_int32_t dlfs_interleave; /* 364: segment interleave */ 702 u_int32_t dlfs_ident; /* 368: per-fs identifier */ 703 u_int32_t dlfs_fsbtodb; /* 372: fsbtodb abd dbtodsb shift constant */ 704 u_int32_t dlfs_resvseg; /* 376: segments reserved for the cleaner */ 705 int8_t dlfs_pad[128]; /* 380: round to 512 bytes */ 706 /* Checksum -- last valid disk field. */ 707 u_int32_t dlfs_cksum; /* 508: checksum for superblock checking */ 708 }; 709 710 /* Type used for the inode bitmap */ 711 typedef u_int32_t lfs_bm_t; 712 713 /* 714 * Linked list of segments whose byte count needs updating following a 715 * file truncation. 716 */ 717 struct segdelta { 718 long segnum; 719 size_t num; 720 LIST_ENTRY(segdelta) list; 721 }; 722 723 /* 724 * In-memory super block. 725 */ 726 struct lfs { 727 struct dlfs lfs_dlfs; /* on-disk parameters */ 728 #define lfs_magic lfs_dlfs.dlfs_magic 729 #define lfs_version lfs_dlfs.dlfs_version 730 #define lfs_size lfs_dlfs.dlfs_size 731 #define lfs_ssize lfs_dlfs.dlfs_ssize 732 #define lfs_dsize lfs_dlfs.dlfs_dsize 733 #define lfs_bsize lfs_dlfs.dlfs_bsize 734 #define lfs_fsize lfs_dlfs.dlfs_fsize 735 #define lfs_frag lfs_dlfs.dlfs_frag 736 #define lfs_freehd lfs_dlfs.dlfs_freehd 737 #define lfs_bfree lfs_dlfs.dlfs_bfree 738 #define lfs_nfiles lfs_dlfs.dlfs_nfiles 739 #define lfs_avail lfs_dlfs.dlfs_avail 740 #define lfs_uinodes lfs_dlfs.dlfs_uinodes 741 #define lfs_idaddr lfs_dlfs.dlfs_idaddr 742 #define lfs_ifile lfs_dlfs.dlfs_ifile 743 #define lfs_lastseg lfs_dlfs.dlfs_lastseg 744 #define lfs_nextseg lfs_dlfs.dlfs_nextseg 745 #define lfs_curseg lfs_dlfs.dlfs_curseg 746 #define lfs_offset lfs_dlfs.dlfs_offset 747 #define lfs_lastpseg lfs_dlfs.dlfs_lastpseg 748 #define lfs_otstamp lfs_dlfs.dlfs_inopf 749 #define lfs_inopf lfs_dlfs.dlfs_inopf 750 #define lfs_minfree lfs_dlfs.dlfs_minfree 751 #define lfs_maxfilesize lfs_dlfs.dlfs_maxfilesize 752 #define lfs_fsbpseg lfs_dlfs.dlfs_fsbpseg 753 #define lfs_inopb lfs_dlfs.dlfs_inopb 754 #define lfs_ifpb lfs_dlfs.dlfs_ifpb 755 #define lfs_sepb lfs_dlfs.dlfs_sepb 756 #define lfs_nindir lfs_dlfs.dlfs_nindir 757 #define lfs_nseg lfs_dlfs.dlfs_nseg 758 #define lfs_nspf lfs_dlfs.dlfs_nspf 759 #define lfs_cleansz lfs_dlfs.dlfs_cleansz 760 #define lfs_segtabsz lfs_dlfs.dlfs_segtabsz 761 #define lfs_segmask lfs_dlfs.dlfs_segmask 762 #define lfs_segshift lfs_dlfs.dlfs_segshift 763 #define lfs_bmask lfs_dlfs.dlfs_bmask 764 #define lfs_bshift lfs_dlfs.dlfs_bshift 765 #define lfs_ffmask lfs_dlfs.dlfs_ffmask 766 #define lfs_ffshift lfs_dlfs.dlfs_ffshift 767 #define lfs_fbmask lfs_dlfs.dlfs_fbmask 768 #define lfs_fbshift lfs_dlfs.dlfs_fbshift 769 #define lfs_blktodb lfs_dlfs.dlfs_blktodb 770 #define lfs_fsbtodb lfs_dlfs.dlfs_fsbtodb 771 #define lfs_sushift lfs_dlfs.dlfs_sushift 772 #define lfs_maxsymlinklen lfs_dlfs.dlfs_maxsymlinklen 773 #define lfs_sboffs lfs_dlfs.dlfs_sboffs 774 #define lfs_cksum lfs_dlfs.dlfs_cksum 775 #define lfs_pflags lfs_dlfs.dlfs_pflags 776 #define lfs_fsmnt lfs_dlfs.dlfs_fsmnt 777 #define lfs_nclean lfs_dlfs.dlfs_nclean 778 #define lfs_dmeta lfs_dlfs.dlfs_dmeta 779 #define lfs_minfreeseg lfs_dlfs.dlfs_minfreeseg 780 #define lfs_sumsize lfs_dlfs.dlfs_sumsize 781 #define lfs_serial lfs_dlfs.dlfs_serial 782 #define lfs_ibsize lfs_dlfs.dlfs_ibsize 783 #define lfs_start lfs_dlfs.dlfs_start 784 #define lfs_tstamp lfs_dlfs.dlfs_tstamp 785 #define lfs_inodefmt lfs_dlfs.dlfs_inodefmt 786 #define lfs_interleave lfs_dlfs.dlfs_interleave 787 #define lfs_ident lfs_dlfs.dlfs_ident 788 #define lfs_resvseg lfs_dlfs.dlfs_resvseg 789 790 /* These fields are set at mount time and are meaningless on disk. */ 791 struct segment *lfs_sp; /* current segment being written */ 792 struct vnode *lfs_ivnode; /* vnode for the ifile */ 793 u_int32_t lfs_seglock; /* single-thread the segment writer */ 794 pid_t lfs_lockpid; /* pid of lock holder */ 795 lwpid_t lfs_locklwp; /* lwp of lock holder */ 796 u_int32_t lfs_iocount; /* number of ios pending */ 797 u_int32_t lfs_writer; /* don't allow any dirops to start */ 798 u_int32_t lfs_dirops; /* count of active directory ops */ 799 u_int32_t lfs_dirvcount; /* count of VDIROP nodes in this fs */ 800 u_int32_t lfs_doifile; /* Write ifile blocks on next write */ 801 u_int32_t lfs_nactive; /* Number of segments since last ckp */ 802 int8_t lfs_fmod; /* super block modified flag */ 803 int8_t lfs_ronly; /* mounted read-only flag */ 804 #define LFS_NOTYET 0x01 805 #define LFS_IFDIRTY 0x02 806 #define LFS_WARNED 0x04 807 #define LFS_UNDIROP 0x08 808 int8_t lfs_flags; /* currently unused flag */ 809 u_int16_t lfs_activesb; /* toggle between superblocks */ 810 daddr_t lfs_sbactive; /* disk address of current sb write */ 811 struct vnode *lfs_flushvp; /* vnode being flushed */ 812 int lfs_flushvp_fakevref; /* fake vref count for flushvp */ 813 struct vnode *lfs_unlockvp; /* being inactivated in lfs_segunlock */ 814 u_int32_t lfs_diropwait; /* # procs waiting on dirop flush */ 815 size_t lfs_devbsize; /* Device block size */ 816 size_t lfs_devbshift; /* Device block shift */ 817 krwlock_t lfs_fraglock; 818 krwlock_t lfs_iflock; /* Ifile lock */ 819 kcondvar_t lfs_stopcv; /* Wrap lock */ 820 struct lwp *lfs_stoplwp; 821 pid_t lfs_rfpid; /* Process ID of roll-forward agent */ 822 int lfs_nadirop; /* number of active dirop nodes */ 823 long lfs_ravail; /* blocks pre-reserved for writing */ 824 long lfs_favail; /* blocks pre-reserved for writing */ 825 res_t *lfs_resblk; /* Reserved memory for pageout */ 826 TAILQ_HEAD(, inode) lfs_dchainhd; /* dirop vnodes */ 827 TAILQ_HEAD(, inode) lfs_pchainhd; /* paging vnodes */ 828 #define LFS_RESHASH_WIDTH 17 829 LIST_HEAD(, lfs_res_blk) lfs_reshash[LFS_RESHASH_WIDTH]; 830 int lfs_pdflush; /* pagedaemon wants us to flush */ 831 u_int32_t **lfs_suflags; /* Segment use flags */ 832 #ifdef _KERNEL 833 struct pool lfs_clpool; /* Pool for struct lfs_cluster */ 834 struct pool lfs_bpppool; /* Pool for bpp */ 835 struct pool lfs_segpool; /* Pool for struct segment */ 836 #endif /* _KERNEL */ 837 #define LFS_MAX_CLEANIND 64 838 int32_t lfs_cleanint[LFS_MAX_CLEANIND]; /* Active cleaning intervals */ 839 int lfs_cleanind; /* Index into intervals */ 840 int lfs_sleepers; /* # procs sleeping this fs */ 841 int lfs_pages; /* dirty pages blaming this fs */ 842 lfs_bm_t *lfs_ino_bitmap; /* Inuse inodes bitmap */ 843 int lfs_nowrap; /* Suspend log wrap */ 844 int lfs_wrappass; /* Allow first log wrap requester to pass */ 845 int lfs_wrapstatus; /* Wrap status */ 846 LIST_HEAD(, segdelta) lfs_segdhd; /* List of pending trunc accounting events */ 847 }; 848 849 /* NINDIR is the number of indirects in a file system block. */ 850 #define NINDIR(fs) ((fs)->lfs_nindir) 851 852 /* INOPB is the number of inodes in a secondary storage block. */ 853 #define INOPB(fs) ((fs)->lfs_inopb) 854 /* INOPF is the number of inodes in a fragment. */ 855 #define INOPF(fs) ((fs)->lfs_inopf) 856 857 #define blksize(fs, ip, lbn) \ 858 (((lbn) >= NDADDR || (ip)->i_ffs1_size >= ((lbn) + 1) << (fs)->lfs_bshift) \ 859 ? (fs)->lfs_bsize \ 860 : (fragroundup(fs, blkoff(fs, (ip)->i_ffs1_size)))) 861 #define blkoff(fs, loc) ((int)((loc) & (fs)->lfs_bmask)) 862 #define fragoff(fs, loc) /* calculates (loc % fs->lfs_fsize) */ \ 863 ((int)((loc) & (fs)->lfs_ffmask)) 864 #define fsbtodb(fs, b) ((b) << (fs)->lfs_fsbtodb) 865 #define dbtofsb(fs, b) ((b) >> (fs)->lfs_fsbtodb) 866 #define fragstodb(fs, b) ((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift)) 867 #define dbtofrags(fs, b) ((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift)) 868 #define lblkno(fs, loc) ((loc) >> (fs)->lfs_bshift) 869 #define lblktosize(fs, blk) ((blk) << (fs)->lfs_bshift) 870 /* Same as above, but named like dbtob(), btodb() */ 871 #define fsbtob(fs, b) ((b) << ((fs)->lfs_bshift - \ 872 (fs)->lfs_blktodb + (fs)->lfs_fsbtodb)) 873 #define btofsb(fs, b) ((b) >> ((fs)->lfs_bshift - \ 874 (fs)->lfs_blktodb + (fs)->lfs_fsbtodb)) 875 #define fsbtofrags(fs, b) ((b) >> ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \ 876 (fs)->lfs_fsbtodb)) 877 #define fragstofsb(fs, b) ((b) << ((fs)->lfs_blktodb - (fs)->lfs_fbshift - \ 878 (fs)->lfs_fsbtodb)) 879 #define btofrags(fs, b) ((b) >> (fs)->lfs_ffshift) 880 #define numfrags(fs, loc) /* calculates (loc / fs->lfs_fsize) */ \ 881 ((loc) >> (fs)->lfs_ffshift) 882 #define blkroundup(fs, size) /* calculates roundup(size, fs->lfs_bsize) */ \ 883 ((off_t)(((size) + (fs)->lfs_bmask) & (~(fs)->lfs_bmask))) 884 #define fragroundup(fs, size) /* calculates roundup(size, fs->lfs_fsize) */ \ 885 ((off_t)(((size) + (fs)->lfs_ffmask) & (~(fs)->lfs_ffmask))) 886 #define fragstoblks(fs, frags) /* calculates (frags / fs->lfs_frag) */ \ 887 ((frags) >> (fs)->lfs_fbshift) 888 #define blkstofrags(fs, blks) /* calculates (blks * fs->lfs_frag) */ \ 889 ((blks) << (fs)->lfs_fbshift) 890 #define fragnum(fs, fsb) /* calculates (fsb % fs->lfs_frag) */ \ 891 ((fsb) & ((fs)->lfs_frag - 1)) 892 #define blknum(fs, fsb) /* calculates rounddown(fsb, fs->lfs_frag) */ \ 893 ((fsb) &~ ((fs)->lfs_frag - 1)) 894 #define dblksize(fs, dp, lbn) \ 895 (((lbn) >= NDADDR || (dp)->di_size >= ((lbn) + 1) << (fs)->lfs_bshift)\ 896 ? (fs)->lfs_bsize \ 897 : (fragroundup(fs, blkoff(fs, (dp)->di_size)))) 898 899 #define segsize(fs) ((fs)->lfs_version == 1 ? \ 900 lblktosize((fs), (fs)->lfs_ssize) : \ 901 (fs)->lfs_ssize) 902 #define segtod(fs, seg) (((fs)->lfs_version == 1 ? \ 903 (fs)->lfs_ssize << (fs)->lfs_blktodb : \ 904 btofsb((fs), (fs)->lfs_ssize)) * (seg)) 905 #define dtosn(fs, daddr) /* block address to segment number */ \ 906 ((uint32_t)(((daddr) - (fs)->lfs_start) / segtod((fs), 1))) 907 #define sntod(fs, sn) /* segment number to disk address */ \ 908 ((daddr_t)(segtod((fs), (sn)) + (fs)->lfs_start)) 909 910 /* 911 * Structures used by lfs_bmapv and lfs_markv to communicate information 912 * about inodes and data blocks. 913 */ 914 typedef struct block_info { 915 u_int32_t bi_inode; /* inode # */ 916 int32_t bi_lbn; /* logical block w/in file */ 917 int32_t bi_daddr; /* disk address of block */ 918 u_int64_t bi_segcreate; /* origin segment create time */ 919 int bi_version; /* file version number */ 920 void *bi_bp; /* data buffer */ 921 int bi_size; /* size of the block (if fragment) */ 922 } BLOCK_INFO; 923 924 /* Compatibility for 1.5 binaries */ 925 typedef struct block_info_15 { 926 u_int32_t bi_inode; /* inode # */ 927 int32_t bi_lbn; /* logical block w/in file */ 928 int32_t bi_daddr; /* disk address of block */ 929 u_int32_t bi_segcreate; /* origin segment create time */ 930 int bi_version; /* file version number */ 931 void *bi_bp; /* data buffer */ 932 int bi_size; /* size of the block (if fragment) */ 933 } BLOCK_INFO_15; 934 935 /* In-memory description of a segment about to be written. */ 936 struct segment { 937 struct lfs *fs; /* file system pointer */ 938 struct buf **bpp; /* pointer to buffer array */ 939 struct buf **cbpp; /* pointer to next available bp */ 940 struct buf **start_bpp; /* pointer to first bp in this set */ 941 struct buf *ibp; /* buffer pointer to inode page */ 942 struct ufs1_dinode *idp; /* pointer to ifile dinode */ 943 struct finfo *fip; /* current fileinfo pointer */ 944 struct vnode *vp; /* vnode being gathered */ 945 void *segsum; /* segment summary info */ 946 u_int32_t ninodes; /* number of inodes in this segment */ 947 int32_t seg_bytes_left; /* bytes left in segment */ 948 int32_t sum_bytes_left; /* bytes left in summary block */ 949 u_int32_t seg_number; /* number of this segment */ 950 int32_t *start_lbp; /* beginning lbn for this set */ 951 952 #define SEGM_CKP 0x01 /* doing a checkpoint */ 953 #define SEGM_CLEAN 0x02 /* cleaner call; don't sort */ 954 #define SEGM_SYNC 0x04 /* wait for segment */ 955 #define SEGM_PROT 0x08 /* don't inactivate at segunlock */ 956 #define SEGM_PAGEDAEMON 0x10 /* pagedaemon called us */ 957 #define SEGM_WRITERD 0x20 /* LFS writed called us */ 958 #define SEGM_FORCE_CKP 0x40 /* Force checkpoint right away */ 959 u_int16_t seg_flags; /* run-time flags for this segment */ 960 u_int32_t seg_iocount; /* number of ios pending */ 961 int ndupino; /* number of duplicate inodes */ 962 }; 963 964 #ifdef _KERNEL 965 struct lfs_cluster { 966 size_t bufsize; /* Size of kept data */ 967 struct buf **bpp; /* Array of kept buffers */ 968 int bufcount; /* Number of kept buffers */ 969 #define LFS_CL_MALLOC 0x00000001 970 #define LFS_CL_SHIFT 0x00000002 971 #define LFS_CL_SYNC 0x00000004 972 u_int32_t flags; /* Flags */ 973 struct lfs *fs; /* LFS that this belongs to */ 974 struct segment *seg; /* Segment structure, for LFS_CL_SYNC */ 975 }; 976 977 /* 978 * Splay tree containing block numbers allocated through lfs_balloc. 979 */ 980 struct lbnentry { 981 SPLAY_ENTRY(lbnentry) entry; 982 daddr_t lbn; 983 }; 984 #endif /* _KERNEL */ 985 986 /* 987 * LFS inode extensions. 988 */ 989 struct lfs_inode_ext { 990 off_t lfs_osize; /* size of file on disk */ 991 u_int32_t lfs_effnblocks; /* number of blocks when i/o completes */ 992 size_t lfs_fragsize[NDADDR]; /* size of on-disk direct blocks */ 993 TAILQ_ENTRY(inode) lfs_dchain; /* Dirop chain. */ 994 TAILQ_ENTRY(inode) lfs_pchain; /* Paging chain. */ 995 #define LFSI_NO_GOP_WRITE 0x01 996 #define LFSI_DELETED 0x02 997 #define LFSI_WRAPBLOCK 0x04 998 #define LFSI_WRAPWAIT 0x08 999 u_int32_t lfs_iflags; /* Inode flags */ 1000 daddr_t lfs_hiblk; /* Highest lbn held by inode */ 1001 #ifdef _KERNEL 1002 SPLAY_HEAD(lfs_splay, lbnentry) lfs_lbtree; /* Tree of balloc'd lbns */ 1003 int lfs_nbtree; /* Size of tree */ 1004 LIST_HEAD(, segdelta) lfs_segdhd; 1005 #endif 1006 int16_t lfs_odnlink; /* on-disk nlink count for cleaner */ 1007 }; 1008 #define i_lfs_osize inode_ext.lfs->lfs_osize 1009 #define i_lfs_effnblks inode_ext.lfs->lfs_effnblocks 1010 #define i_lfs_fragsize inode_ext.lfs->lfs_fragsize 1011 #define i_lfs_dchain inode_ext.lfs->lfs_dchain 1012 #define i_lfs_pchain inode_ext.lfs->lfs_pchain 1013 #define i_lfs_iflags inode_ext.lfs->lfs_iflags 1014 #define i_lfs_hiblk inode_ext.lfs->lfs_hiblk 1015 #define i_lfs_lbtree inode_ext.lfs->lfs_lbtree 1016 #define i_lfs_nbtree inode_ext.lfs->lfs_nbtree 1017 #define i_lfs_segdhd inode_ext.lfs->lfs_segdhd 1018 #define i_lfs_odnlink inode_ext.lfs->lfs_odnlink 1019 1020 /* 1021 * Macros for determining free space on the disk, with the variable metadata 1022 * of segment summaries and inode blocks taken into account. 1023 */ 1024 /* Estimate number of clean blocks not available for writing */ 1025 #define LFS_EST_CMETA(F) (int32_t)((((F)->lfs_dmeta * \ 1026 (int64_t)(F)->lfs_nclean) / \ 1027 ((F)->lfs_nseg - (F)->lfs_nclean))) 1028 1029 /* Estimate total size of the disk not including metadata */ 1030 #define LFS_EST_NONMETA(F) ((F)->lfs_dsize - (F)->lfs_dmeta - LFS_EST_CMETA(F)) 1031 1032 /* Estimate number of blocks actually available for writing */ 1033 #define LFS_EST_BFREE(F) ((F)->lfs_bfree > LFS_EST_CMETA(F) ? \ 1034 (F)->lfs_bfree - LFS_EST_CMETA(F) : 0) 1035 1036 /* Amount of non-meta space not available to mortal man */ 1037 #define LFS_EST_RSVD(F) (int32_t)((LFS_EST_NONMETA(F) * \ 1038 (u_int64_t)(F)->lfs_minfree) / \ 1039 100) 1040 1041 /* Can credential C write BB blocks */ 1042 #define ISSPACE(F, BB, C) \ 1043 ((((C) == NOCRED || kauth_cred_geteuid(C) == 0) && \ 1044 LFS_EST_BFREE(F) >= (BB)) || \ 1045 (kauth_cred_geteuid(C) != 0 && IS_FREESPACE(F, BB))) 1046 1047 /* Can an ordinary user write BB blocks */ 1048 #define IS_FREESPACE(F, BB) \ 1049 (LFS_EST_BFREE(F) >= (BB) + LFS_EST_RSVD(F)) 1050 1051 /* 1052 * The minimum number of blocks to create a new inode. This is: 1053 * directory direct block (1) + NIADDR indirect blocks + inode block (1) + 1054 * ifile direct block (1) + NIADDR indirect blocks = 3 + 2 * NIADDR blocks. 1055 */ 1056 #define LFS_NRESERVE(F) (btofsb((F), (2 * NIADDR + 3) << (F)->lfs_bshift)) 1057 1058 /* Statistics Counters */ 1059 struct lfs_stats { /* Must match sysctl list in lfs_vfsops.h ! */ 1060 u_int segsused; 1061 u_int psegwrites; 1062 u_int psyncwrites; 1063 u_int pcleanwrites; 1064 u_int blocktot; 1065 u_int cleanblocks; 1066 u_int ncheckpoints; 1067 u_int nwrites; 1068 u_int nsync_writes; 1069 u_int wait_exceeded; 1070 u_int write_exceeded; 1071 u_int flush_invoked; 1072 u_int vflush_invoked; 1073 u_int clean_inlocked; 1074 u_int clean_vnlocked; 1075 u_int segs_reclaimed; 1076 }; 1077 #ifdef _KERNEL 1078 extern struct lfs_stats lfs_stats; 1079 #endif 1080 1081 /* Fcntls to take the place of the lfs syscalls */ 1082 struct lfs_fcntl_markv { 1083 BLOCK_INFO *blkiov; /* blocks to relocate */ 1084 int blkcnt; /* number of blocks */ 1085 }; 1086 1087 #define LFCNSEGWAITALL _FCNR_FSPRIV('L', 14, struct timeval) 1088 #define LFCNSEGWAIT _FCNR_FSPRIV('L', 15, struct timeval) 1089 #define LFCNBMAPV _FCNRW_FSPRIV('L', 2, struct lfs_fcntl_markv) 1090 #define LFCNMARKV _FCNRW_FSPRIV('L', 3, struct lfs_fcntl_markv) 1091 #define LFCNRECLAIM _FCNO_FSPRIV('L', 4) 1092 1093 struct lfs_fhandle { 1094 char space[28]; /* FHANDLE_SIZE_COMPAT (but used from userland too) */ 1095 }; 1096 #define LFCNREWIND _FCNR_FSPRIV('L', 6, int) 1097 #define LFCNINVAL _FCNR_FSPRIV('L', 7, int) 1098 #define LFCNRESIZE _FCNR_FSPRIV('L', 8, int) 1099 #define LFCNWRAPSTOP _FCNR_FSPRIV('L', 9, int) 1100 #define LFCNWRAPGO _FCNR_FSPRIV('L', 10, int) 1101 #define LFCNIFILEFH _FCNW_FSPRIV('L', 11, struct lfs_fhandle) 1102 #define LFCNWRAPPASS _FCNR_FSPRIV('L', 12, int) 1103 # define LFS_WRAP_GOING 0x0 1104 # define LFS_WRAP_WAITING 0x1 1105 #define LFCNWRAPSTATUS _FCNW_FSPRIV('L', 13, int) 1106 /* Compat */ 1107 #ifdef COMPAT_20 1108 #define LFCNSEGWAITALL_COMPAT _FCNW_FSPRIV('L', 0, struct timeval50) 1109 #define LFCNSEGWAIT_COMPAT _FCNW_FSPRIV('L', 1, struct timeval50) 1110 #endif 1111 #ifdef COMPAT_30 1112 #define LFCNIFILEFH_COMPAT _FCNW_FSPRIV('L', 5, struct lfs_fhandle) 1113 #endif 1114 #ifdef COMPAT_40 1115 #define LFCNIFILEFH_COMPAT2 _FCN_FSPRIV(F_FSOUT, 'L', 11, 32) 1116 #define LFCNWRAPSTOP_COMPAT _FCNO_FSPRIV('L', 9) 1117 #define LFCNWRAPGO_COMPAT _FCNO_FSPRIV('L', 10) 1118 #endif 1119 #ifdef COMPAT_50 1120 #define LFCNSEGWAITALL_COMPAT_50 _FCNR_FSPRIV('L', 0, struct timeval50) 1121 #define LFCNSEGWAIT_COMPAT_50 _FCNR_FSPRIV('L', 1, struct timeval50) 1122 #endif 1123 1124 #ifdef _KERNEL 1125 /* XXX MP */ 1126 #define LFS_SEGLOCK_HELD(fs) \ 1127 ((fs)->lfs_seglock != 0 && \ 1128 (fs)->lfs_lockpid == curproc->p_pid && \ 1129 (fs)->lfs_locklwp == curlwp->l_lid) 1130 #endif /* _KERNEL */ 1131 1132 /* Debug segment lock */ 1133 #ifdef notyet 1134 # define ASSERT_SEGLOCK(fs) KASSERT(LFS_SEGLOCK_HELD(fs)) 1135 # define ASSERT_NO_SEGLOCK(fs) KASSERT(!LFS_SEGLOCK_HELD(fs)) 1136 # define ASSERT_DUNNO_SEGLOCK(fs) 1137 # define ASSERT_MAYBE_SEGLOCK(fs) 1138 #else /* !notyet */ 1139 # define ASSERT_DUNNO_SEGLOCK(fs) \ 1140 DLOG((DLOG_SEG, "lfs func %s seglock wrong (%d)\n", __func__, \ 1141 LFS_SEGLOCK_HELD(fs))) 1142 # define ASSERT_SEGLOCK(fs) do { \ 1143 if (!LFS_SEGLOCK_HELD(fs)) { \ 1144 DLOG((DLOG_SEG, "lfs func %s seglock wrong (0)\n", __func__)); \ 1145 } \ 1146 } while(0) 1147 # define ASSERT_NO_SEGLOCK(fs) do { \ 1148 if (LFS_SEGLOCK_HELD(fs)) { \ 1149 DLOG((DLOG_SEG, "lfs func %s seglock wrong (1)\n", __func__)); \ 1150 } \ 1151 } while(0) 1152 # define ASSERT_MAYBE_SEGLOCK(x) 1153 #endif /* !notyet */ 1154 1155 __BEGIN_DECLS 1156 void lfs_itimes(struct inode *, const struct timespec *, 1157 const struct timespec *, const struct timespec *); 1158 __END_DECLS 1159 1160 #endif /* !_UFS_LFS_LFS_H_ */ 1161
Indexes created Mon Sep 29 21:09:56 GMT 2025