lfs_syscalls.c revision 1.18
1 /* $NetBSD: lfs_syscalls.c,v 1.18 1998/06/24 20:58:48 sommerfe Exp $ */ 2 3 /*- 4 * Copyright (c) 1991, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)lfs_syscalls.c 8.10 (Berkeley) 5/14/95 36 */ 37 38 #if defined(_KERNEL) && !defined(_LKM) 39 #include "fs_lfs.h" /* for prototypes in syscallargs.h */ 40 #endif 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/proc.h> 45 #include <sys/buf.h> 46 #include <sys/mount.h> 47 #include <sys/vnode.h> 48 #include <sys/malloc.h> 49 #include <sys/kernel.h> 50 51 #include <sys/syscallargs.h> 52 53 #include <ufs/ufs/quota.h> 54 #include <ufs/ufs/inode.h> 55 #include <ufs/ufs/ufsmount.h> 56 #include <ufs/ufs/ufs_extern.h> 57 58 #include <ufs/lfs/lfs.h> 59 #include <ufs/lfs/lfs_extern.h> 60 61 #define BUMP_FIP(SP) \ 62 (SP)->fip = (FINFO *) (&(SP)->fip->fi_blocks[(SP)->fip->fi_nblocks]) 63 64 #define INC_FINFO(SP) ++((SEGSUM *)((SP)->segsum))->ss_nfinfo 65 #define DEC_FINFO(SP) --((SEGSUM *)((SP)->segsum))->ss_nfinfo 66 67 /* 68 * Before committing to add something to a segment summary, make sure there 69 * is enough room. S is the bytes added to the summary. 70 */ 71 #define CHECK_SEG(s) \ 72 if (sp->sum_bytes_left < (s)) { \ 73 (void) lfs_writeseg(fs, sp); \ 74 } 75 struct buf *lfs_fakebuf __P((struct vnode *, int, size_t, caddr_t)); 76 77 int debug_cleaner = 0; 78 int clean_vnlocked = 0; 79 int clean_inlocked = 0; 80 81 /* 82 * lfs_markv: 83 * 84 * This will mark inodes and blocks dirty, so they are written into the log. 85 * It will block until all the blocks have been written. The segment create 86 * time passed in the block_info and inode_info structures is used to decide 87 * if the data is valid for each block (in case some process dirtied a block 88 * or inode that is being cleaned between the determination that a block is 89 * live and the lfs_markv call). 90 * 91 * 0 on success 92 * -1/errno is return on error. 93 */ 94 int 95 lfs_markv(p, v, retval) 96 struct proc *p; 97 void *v; 98 register_t *retval; 99 { 100 struct lfs_markv_args /* { 101 syscallarg(fsid_t *) fsidp; 102 syscallarg(struct block_info *) blkiov; 103 syscallarg(int) blkcnt; 104 } */ *uap = v; 105 struct segment *sp; 106 BLOCK_INFO *blkp; 107 IFILE *ifp; 108 struct buf *bp, **bpp; 109 struct inode *ip = NULL; 110 struct lfs *fs; 111 struct mount *mntp; 112 struct vnode *vp; 113 fsid_t fsid; 114 void *start; 115 ino_t lastino; 116 ufs_daddr_t b_daddr, v_daddr; 117 u_long bsize; 118 int cnt, error; 119 120 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 121 return (error); 122 123 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 124 return (error); 125 if ((mntp = vfs_getvfs(&fsid)) == NULL) 126 return (EINVAL); 127 128 cnt = SCARG(uap, blkcnt); 129 start = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK); 130 error = copyin(SCARG(uap, blkiov), start, cnt * sizeof(BLOCK_INFO)); 131 if (error) 132 goto err1; 133 134 /* Mark blocks/inodes dirty. */ 135 fs = VFSTOUFS(mntp)->um_lfs; 136 bsize = fs->lfs_bsize; 137 error = 0; 138 139 lfs_seglock(fs, SEGM_SYNC | SEGM_CLEAN); 140 sp = fs->lfs_sp; 141 for (v_daddr = LFS_UNUSED_DADDR, lastino = LFS_UNUSED_INUM, 142 blkp = start; cnt--; ++blkp) { 143 /* 144 * Get the IFILE entry (only once) and see if the file still 145 * exists. 146 */ 147 if (lastino != blkp->bi_inode) { 148 if (lastino != LFS_UNUSED_INUM) { 149 /* Finish up last file */ 150 if (sp->fip->fi_nblocks == 0) { 151 DEC_FINFO(sp); 152 sp->sum_bytes_left += 153 sizeof(FINFO) - sizeof(ufs_daddr_t); 154 } else { 155 lfs_updatemeta(sp); 156 BUMP_FIP(sp); 157 } 158 159 lfs_writeinode(fs, sp, ip); 160 lfs_vunref(vp); 161 } 162 163 /* Start a new file */ 164 CHECK_SEG(sizeof(FINFO)); 165 sp->sum_bytes_left -= sizeof(FINFO) - sizeof(ufs_daddr_t); 166 INC_FINFO(sp); 167 sp->start_lbp = &sp->fip->fi_blocks[0]; 168 sp->vp = NULL; 169 sp->fip->fi_version = blkp->bi_version; 170 sp->fip->fi_nblocks = 0; 171 sp->fip->fi_ino = blkp->bi_inode; 172 lastino = blkp->bi_inode; 173 if (blkp->bi_inode == LFS_IFILE_INUM) 174 v_daddr = fs->lfs_idaddr; 175 else { 176 LFS_IENTRY(ifp, fs, blkp->bi_inode, bp); 177 v_daddr = ifp->if_daddr; 178 brelse(bp); 179 } 180 if (v_daddr == LFS_UNUSED_DADDR) 181 continue; 182 183 /* Get the vnode/inode. */ 184 if (lfs_fastvget(mntp, blkp->bi_inode, v_daddr, &vp, 185 blkp->bi_lbn == LFS_UNUSED_LBN ? 186 blkp->bi_bp : NULL)) { 187 #ifdef DIAGNOSTIC 188 printf("lfs_markv: VFS_VGET failed (%d)\n", 189 blkp->bi_inode); 190 panic("lfs_markv VFS_VGET FAILED"); 191 #endif 192 lastino = LFS_UNUSED_INUM; 193 v_daddr = LFS_UNUSED_DADDR; 194 continue; 195 } 196 sp->vp = vp; 197 ip = VTOI(vp); 198 } else if (v_daddr == LFS_UNUSED_DADDR) 199 continue; 200 201 /* If this BLOCK_INFO didn't contain a block, keep going. */ 202 if (blkp->bi_lbn == LFS_UNUSED_LBN) 203 continue; 204 if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &b_daddr, NULL) || 205 b_daddr != blkp->bi_daddr) 206 continue; 207 /* 208 * If we got to here, then we are keeping the block. If it 209 * is an indirect block, we want to actually put it in the 210 * buffer cache so that it can be updated in the finish_meta 211 * section. If it's not, we need to allocate a fake buffer 212 * so that writeseg can perform the copyin and write the buffer. 213 */ 214 if (blkp->bi_lbn >= 0) /* Data Block */ 215 bp = lfs_fakebuf(vp, blkp->bi_lbn, bsize, 216 blkp->bi_bp); 217 else { 218 bp = getblk(vp, blkp->bi_lbn, bsize, 0, 0); 219 if (!(bp->b_flags & (B_DELWRI | B_DONE | B_CACHE)) && 220 (error = copyin(blkp->bi_bp, bp->b_data, 221 blkp->bi_size))) 222 goto err2; 223 if ((error = VOP_BWRITE(bp)) != 0) 224 goto err2; 225 } 226 while (lfs_gatherblock(sp, bp, NULL)); 227 } 228 if (sp->vp) { 229 if (sp->fip->fi_nblocks == 0) { 230 DEC_FINFO(sp); 231 sp->sum_bytes_left += 232 sizeof(FINFO) - sizeof(ufs_daddr_t); 233 } else 234 lfs_updatemeta(sp); 235 236 lfs_writeinode(fs, sp, ip); 237 lfs_vunref(vp); 238 } 239 (void) lfs_writeseg(fs, sp); 240 lfs_segunlock(fs); 241 free(start, M_SEGMENT); 242 return (error); 243 244 /* 245 * XXX 246 * If we come in to error 2, we might have indirect blocks that were 247 * updated and now have bad block pointers. I don't know what to do 248 * about this. 249 */ 250 251 err2: lfs_vunref(vp); 252 /* Free up fakebuffers */ 253 for (bpp = --sp->cbpp; bpp >= sp->bpp; --bpp) 254 if ((*bpp)->b_flags & B_CALL) { 255 brelvp(*bpp); 256 free(*bpp, M_SEGMENT); 257 } else 258 brelse(*bpp); 259 lfs_segunlock(fs); 260 err1: 261 free(start, M_SEGMENT); 262 return (error); 263 } 264 265 /* 266 * lfs_bmapv: 267 * 268 * This will fill in the current disk address for arrays of blocks. 269 * 270 * 0 on success 271 * -1/errno is return on error. 272 */ 273 int 274 lfs_bmapv(p, v, retval) 275 struct proc *p; 276 void *v; 277 register_t *retval; 278 { 279 struct lfs_bmapv_args /* { 280 syscallarg(fsid_t *) fsidp; 281 syscallarg(struct block_info *) blkiov; 282 syscallarg(int) blkcnt; 283 } */ *uap = v; 284 BLOCK_INFO *blkp; 285 struct mount *mntp; 286 struct ufsmount *ump; 287 struct vnode *vp; 288 fsid_t fsid; 289 void *start; 290 ufs_daddr_t daddr; 291 int cnt, error, step; 292 293 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 294 return (error); 295 296 error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t)); 297 if (error) 298 return (error); 299 if ((mntp = vfs_getvfs(&fsid)) == NULL) 300 return (EINVAL); 301 302 cnt = SCARG(uap, blkcnt); 303 start = blkp = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK); 304 error = copyin(SCARG(uap, blkiov), blkp, cnt * sizeof(BLOCK_INFO)); 305 if (error) { 306 free(blkp, M_SEGMENT); 307 return (error); 308 } 309 310 for (step = cnt; step--; ++blkp) { 311 if (blkp->bi_lbn == LFS_UNUSED_LBN) 312 continue; 313 /* 314 * A regular call to VFS_VGET could deadlock 315 * here. Instead, we try an unlocked access. 316 */ 317 ump = VFSTOUFS(mntp); 318 if ((vp = 319 ufs_ihashlookup(ump->um_dev, blkp->bi_inode)) != NULL) { 320 if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &daddr, NULL)) 321 daddr = LFS_UNUSED_DADDR; 322 } else if (VFS_VGET(mntp, blkp->bi_inode, &vp)) 323 daddr = LFS_UNUSED_DADDR; 324 else { 325 if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &daddr, NULL)) 326 daddr = LFS_UNUSED_DADDR; 327 vput(vp); 328 } 329 blkp->bi_daddr = daddr; 330 } 331 copyout(start, SCARG(uap, blkiov), cnt * sizeof(BLOCK_INFO)); 332 free(start, M_SEGMENT); 333 return (0); 334 } 335 336 /* 337 * lfs_segclean: 338 * 339 * Mark the segment clean. 340 * 341 * 0 on success 342 * -1/errno is return on error. 343 */ 344 int 345 lfs_segclean(p, v, retval) 346 struct proc *p; 347 void *v; 348 register_t *retval; 349 { 350 struct lfs_segclean_args /* { 351 syscallarg(fsid_t *) fsidp; 352 syscallarg(u_long) segment; 353 } */ *uap = v; 354 CLEANERINFO *cip; 355 SEGUSE *sup; 356 struct buf *bp; 357 struct mount *mntp; 358 struct lfs *fs; 359 fsid_t fsid; 360 int error; 361 362 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 363 return (error); 364 365 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 366 return (error); 367 if ((mntp = vfs_getvfs(&fsid)) == NULL) 368 return (EINVAL); 369 370 fs = VFSTOUFS(mntp)->um_lfs; 371 372 if (datosn(fs, fs->lfs_curseg) == SCARG(uap, segment)) 373 return (EBUSY); 374 375 LFS_SEGENTRY(sup, fs, SCARG(uap, segment), bp); 376 if (sup->su_flags & SEGUSE_ACTIVE) { 377 brelse(bp); 378 return (EBUSY); 379 } 380 fs->lfs_avail += fsbtodb(fs, fs->lfs_ssize) - 1; 381 fs->lfs_bfree += (sup->su_nsums * LFS_SUMMARY_SIZE / DEV_BSIZE) + 382 sup->su_ninos * btodb(fs->lfs_bsize); 383 sup->su_flags &= ~SEGUSE_DIRTY; 384 (void) VOP_BWRITE(bp); 385 386 LFS_CLEANERINFO(cip, fs, bp); 387 ++cip->clean; 388 --cip->dirty; 389 (void) VOP_BWRITE(bp); 390 wakeup(&fs->lfs_avail); 391 return (0); 392 } 393 394 /* 395 * lfs_segwait: 396 * 397 * This will block until a segment in file system fsid is written. A timeout 398 * in milliseconds may be specified which will awake the cleaner automatically. 399 * An fsid of -1 means any file system, and a timeout of 0 means forever. 400 * 401 * 0 on success 402 * 1 on timeout 403 * -1/errno is return on error. 404 */ 405 int 406 lfs_segwait(p, v, retval) 407 struct proc *p; 408 void *v; 409 register_t *retval; 410 { 411 struct lfs_segwait_args /* { 412 syscallarg(fsid_t *) fsidp; 413 syscallarg(struct timeval *) tv; 414 } */ *uap = v; 415 extern int lfs_allclean_wakeup; 416 struct mount *mntp; 417 struct timeval atv; 418 fsid_t fsid; 419 void *addr; 420 u_long timeout; 421 int error, s; 422 423 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) { 424 return (error); 425 } 426 #ifdef WHEN_QUADS_WORK 427 if (error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) 428 return (error); 429 if (fsid == (fsid_t)-1) 430 addr = &lfs_allclean_wakeup; 431 else { 432 if ((mntp = vfs_getvfs(&fsid)) == NULL) 433 return (EINVAL); 434 addr = &VFSTOUFS(mntp)->um_lfs->lfs_nextseg; 435 } 436 #else 437 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 438 return (error); 439 if ((mntp = vfs_getvfs(&fsid)) == NULL) 440 addr = &lfs_allclean_wakeup; 441 else 442 addr = &VFSTOUFS(mntp)->um_lfs->lfs_nextseg; 443 #endif 444 445 if (SCARG(uap, tv)) { 446 error = copyin(SCARG(uap, tv), &atv, sizeof(struct timeval)); 447 if (error) 448 return (error); 449 if (itimerfix(&atv)) 450 return (EINVAL); 451 s = splclock(); 452 timeradd(&atv, &time, &atv); 453 timeout = hzto(&atv); 454 splx(s); 455 } else 456 timeout = 0; 457 458 error = tsleep(addr, PCATCH | PUSER, "segment", timeout); 459 return (error == ERESTART ? EINTR : 0); 460 } 461 462 /* 463 * VFS_VGET call specialized for the cleaner. The cleaner already knows the 464 * daddr from the ifile, so don't look it up again. If the cleaner is 465 * processing IINFO structures, it may have the ondisk inode already, so 466 * don't go retrieving it again. 467 */ 468 int 469 lfs_fastvget(mp, ino, daddr, vpp, dinp) 470 struct mount *mp; 471 ino_t ino; 472 ufs_daddr_t daddr; 473 struct vnode **vpp; 474 struct dinode *dinp; 475 { 476 register struct inode *ip; 477 struct vnode *vp; 478 struct ufsmount *ump; 479 struct buf *bp; 480 dev_t dev; 481 int error; 482 483 ump = VFSTOUFS(mp); 484 dev = ump->um_dev; 485 /* 486 * This is playing fast and loose. Someone may have the inode 487 * locked, in which case they are going to be distinctly unhappy 488 * if we trash something. 489 */ 490 if ((*vpp = ufs_ihashlookup(dev, ino)) != NULL) { 491 lfs_vref(*vpp); 492 if ((*vpp)->v_flag & VXLOCK) 493 clean_vnlocked++; 494 ip = VTOI(*vpp); 495 if (lockstatus(&ip->i_lock)) 496 clean_inlocked++; 497 if (!(ip->i_flag & IN_MODIFIED)) 498 ++ump->um_lfs->lfs_uinodes; 499 ip->i_flag |= IN_MODIFIED; 500 return (0); 501 } 502 503 /* Allocate new vnode/inode. */ 504 if ((error = lfs_vcreate(mp, ino, &vp)) != 0) { 505 *vpp = NULL; 506 return (error); 507 } 508 509 /* 510 * Put it onto its hash chain and lock it so that other requests for 511 * this inode will block if they arrive while we are sleeping waiting 512 * for old data structures to be purged or for the contents of the 513 * disk portion of this inode to be read. 514 */ 515 ip = VTOI(vp); 516 ufs_ihashins(ip); 517 518 /* 519 * XXX 520 * This may not need to be here, logically it should go down with 521 * the i_devvp initialization. 522 * Ask Kirk. 523 */ 524 ip->i_lfs = ump->um_lfs; 525 526 /* Read in the disk contents for the inode, copy into the inode. */ 527 if (dinp) { 528 error = copyin(dinp, &ip->i_din.ffs_din, sizeof(struct dinode)); 529 if (error) 530 return (error); 531 } 532 else { 533 error = bread(ump->um_devvp, daddr, 534 (int)ump->um_lfs->lfs_bsize, NOCRED, &bp); 535 if (error) { 536 /* 537 * The inode does not contain anything useful, so it 538 * would be misleading to leave it on its hash chain. 539 * Iput() will return it to the free list. 540 */ 541 ufs_ihashrem(ip); 542 543 /* Unlock and discard unneeded inode. */ 544 lfs_vunref(vp); 545 brelse(bp); 546 *vpp = NULL; 547 return (error); 548 } 549 ip->i_din.ffs_din = 550 *lfs_ifind(ump->um_lfs, ino, (struct dinode *)bp->b_data); 551 brelse(bp); 552 } 553 554 /* 555 * Initialize the vnode from the inode, check for aliases. In all 556 * cases re-init ip, the underlying vnode/inode may have changed. 557 */ 558 error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp); 559 if (error) { 560 lfs_vunref(vp); 561 *vpp = NULL; 562 return (error); 563 } 564 /* 565 * Finish inode initialization now that aliasing has been resolved. 566 */ 567 ip->i_devvp = ump->um_devvp; 568 ip->i_flag |= IN_MODIFIED; 569 ++ump->um_lfs->lfs_uinodes; 570 VREF(ip->i_devvp); 571 *vpp = vp; 572 return (0); 573 } 574 struct buf * 575 lfs_fakebuf(vp, lbn, size, uaddr) 576 struct vnode *vp; 577 int lbn; 578 size_t size; 579 caddr_t uaddr; 580 { 581 struct buf *bp; 582 583 bp = lfs_newbuf(vp, lbn, 0); 584 bp->b_saveaddr = uaddr; 585 bp->b_bufsize = size; 586 bp->b_bcount = size; 587 bp->b_flags |= B_INVAL; 588 return (bp); 589 } 590
Indexes created Tue Sep 30 11:09:46 GMT 2025